1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Portions Copyright 2011 Martin Matuska 25 * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 26 * Copyright (c) 2014, Joyent, Inc. All rights reserved. 27 * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 29 * Copyright (c) 2013 Steven Hartland. All rights reserved. 30 */ 31 32 /* 33 * ZFS ioctls. 34 * 35 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage 36 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool. 37 * 38 * There are two ways that we handle ioctls: the legacy way where almost 39 * all of the logic is in the ioctl callback, and the new way where most 40 * of the marshalling is handled in the common entry point, zfsdev_ioctl(). 41 * 42 * Non-legacy ioctls should be registered by calling 43 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked 44 * from userland by lzc_ioctl(). 45 * 46 * The registration arguments are as follows: 47 * 48 * const char *name 49 * The name of the ioctl. This is used for history logging. If the 50 * ioctl returns successfully (the callback returns 0), and allow_log 51 * is true, then a history log entry will be recorded with the input & 52 * output nvlists. The log entry can be printed with "zpool history -i". 53 * 54 * zfs_ioc_t ioc 55 * The ioctl request number, which userland will pass to ioctl(2). 56 * The ioctl numbers can change from release to release, because 57 * the caller (libzfs) must be matched to the kernel. 58 * 59 * zfs_secpolicy_func_t *secpolicy 60 * This function will be called before the zfs_ioc_func_t, to 61 * determine if this operation is permitted. It should return EPERM 62 * on failure, and 0 on success. Checks include determining if the 63 * dataset is visible in this zone, and if the user has either all 64 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission 65 * to do this operation on this dataset with "zfs allow". 66 * 67 * zfs_ioc_namecheck_t namecheck 68 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool 69 * name, a dataset name, or nothing. If the name is not well-formed, 70 * the ioctl will fail and the callback will not be called. 71 * Therefore, the callback can assume that the name is well-formed 72 * (e.g. is null-terminated, doesn't have more than one '@' character, 73 * doesn't have invalid characters). 74 * 75 * zfs_ioc_poolcheck_t pool_check 76 * This specifies requirements on the pool state. If the pool does 77 * not meet them (is suspended or is readonly), the ioctl will fail 78 * and the callback will not be called. If any checks are specified 79 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME. 80 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED | 81 * POOL_CHECK_READONLY). 82 * 83 * boolean_t smush_outnvlist 84 * If smush_outnvlist is true, then the output is presumed to be a 85 * list of errors, and it will be "smushed" down to fit into the 86 * caller's buffer, by removing some entries and replacing them with a 87 * single "N_MORE_ERRORS" entry indicating how many were removed. See 88 * nvlist_smush() for details. If smush_outnvlist is false, and the 89 * outnvlist does not fit into the userland-provided buffer, then the 90 * ioctl will fail with ENOMEM. 91 * 92 * zfs_ioc_func_t *func 93 * The callback function that will perform the operation. 94 * 95 * The callback should return 0 on success, or an error number on 96 * failure. If the function fails, the userland ioctl will return -1, 97 * and errno will be set to the callback's return value. The callback 98 * will be called with the following arguments: 99 * 100 * const char *name 101 * The name of the pool or dataset to operate on, from 102 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the 103 * expected type (pool, dataset, or none). 104 * 105 * nvlist_t *innvl 106 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or 107 * NULL if no input nvlist was provided. Changes to this nvlist are 108 * ignored. If the input nvlist could not be deserialized, the 109 * ioctl will fail and the callback will not be called. 110 * 111 * nvlist_t *outnvl 112 * The output nvlist, initially empty. The callback can fill it in, 113 * and it will be returned to userland by serializing it into 114 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization 115 * fails (e.g. because the caller didn't supply a large enough 116 * buffer), then the overall ioctl will fail. See the 117 * 'smush_nvlist' argument above for additional behaviors. 118 * 119 * There are two typical uses of the output nvlist: 120 * - To return state, e.g. property values. In this case, 121 * smush_outnvlist should be false. If the buffer was not large 122 * enough, the caller will reallocate a larger buffer and try 123 * the ioctl again. 124 * 125 * - To return multiple errors from an ioctl which makes on-disk 126 * changes. In this case, smush_outnvlist should be true. 127 * Ioctls which make on-disk modifications should generally not 128 * use the outnvl if they succeed, because the caller can not 129 * distinguish between the operation failing, and 130 * deserialization failing. 131 */ 132 133 #include <sys/types.h> 134 #include <sys/param.h> 135 #include <sys/errno.h> 136 #include <sys/uio.h> 137 #include <sys/buf.h> 138 #include <sys/modctl.h> 139 #include <sys/open.h> 140 #include <sys/file.h> 141 #include <sys/kmem.h> 142 #include <sys/conf.h> 143 #include <sys/cmn_err.h> 144 #include <sys/stat.h> 145 #include <sys/zfs_ioctl.h> 146 #include <sys/zfs_vfsops.h> 147 #include <sys/zfs_znode.h> 148 #include <sys/zap.h> 149 #include <sys/spa.h> 150 #include <sys/spa_impl.h> 151 #include <sys/vdev.h> 152 #include <sys/priv_impl.h> 153 #include <sys/dmu.h> 154 #include <sys/dsl_dir.h> 155 #include <sys/dsl_dataset.h> 156 #include <sys/dsl_prop.h> 157 #include <sys/dsl_deleg.h> 158 #include <sys/dmu_objset.h> 159 #include <sys/dmu_impl.h> 160 #include <sys/dmu_tx.h> 161 #include <sys/ddi.h> 162 #include <sys/sunddi.h> 163 #include <sys/sunldi.h> 164 #include <sys/policy.h> 165 #include <sys/zone.h> 166 #include <sys/nvpair.h> 167 #include <sys/pathname.h> 168 #include <sys/mount.h> 169 #include <sys/sdt.h> 170 #include <sys/fs/zfs.h> 171 #include <sys/zfs_ctldir.h> 172 #include <sys/zfs_dir.h> 173 #include <sys/zfs_onexit.h> 174 #include <sys/zvol.h> 175 #include <sys/dsl_scan.h> 176 #include <sharefs/share.h> 177 #include <sys/dmu_objset.h> 178 #include <sys/dmu_send.h> 179 #include <sys/dsl_destroy.h> 180 #include <sys/dsl_bookmark.h> 181 #include <sys/dsl_userhold.h> 182 #include <sys/zfeature.h> 183 184 #include "zfs_namecheck.h" 185 #include "zfs_prop.h" 186 #include "zfs_deleg.h" 187 #include "zfs_comutil.h" 188 189 extern struct modlfs zfs_modlfs; 190 191 extern void zfs_init(void); 192 extern void zfs_fini(void); 193 194 ldi_ident_t zfs_li = NULL; 195 dev_info_t *zfs_dip; 196 197 uint_t zfs_fsyncer_key; 198 extern uint_t rrw_tsd_key; 199 static uint_t zfs_allow_log_key; 200 201 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *); 202 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *); 203 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *); 204 205 typedef enum { 206 NO_NAME, 207 POOL_NAME, 208 DATASET_NAME 209 } zfs_ioc_namecheck_t; 210 211 typedef enum { 212 POOL_CHECK_NONE = 1 << 0, 213 POOL_CHECK_SUSPENDED = 1 << 1, 214 POOL_CHECK_READONLY = 1 << 2, 215 } zfs_ioc_poolcheck_t; 216 217 typedef struct zfs_ioc_vec { 218 zfs_ioc_legacy_func_t *zvec_legacy_func; 219 zfs_ioc_func_t *zvec_func; 220 zfs_secpolicy_func_t *zvec_secpolicy; 221 zfs_ioc_namecheck_t zvec_namecheck; 222 boolean_t zvec_allow_log; 223 zfs_ioc_poolcheck_t zvec_pool_check; 224 boolean_t zvec_smush_outnvlist; 225 const char *zvec_name; 226 } zfs_ioc_vec_t; 227 228 /* This array is indexed by zfs_userquota_prop_t */ 229 static const char *userquota_perms[] = { 230 ZFS_DELEG_PERM_USERUSED, 231 ZFS_DELEG_PERM_USERQUOTA, 232 ZFS_DELEG_PERM_GROUPUSED, 233 ZFS_DELEG_PERM_GROUPQUOTA, 234 }; 235 236 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc); 237 static int zfs_check_settable(const char *name, nvpair_t *property, 238 cred_t *cr); 239 static int zfs_check_clearable(char *dataset, nvlist_t *props, 240 nvlist_t **errors); 241 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *, 242 boolean_t *); 243 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); 244 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp); 245 246 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature); 247 248 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */ 249 void 250 __dprintf(const char *file, const char *func, int line, const char *fmt, ...) 251 { 252 const char *newfile; 253 char buf[512]; 254 va_list adx; 255 256 /* 257 * Get rid of annoying "../common/" prefix to filename. 258 */ 259 newfile = strrchr(file, '/'); 260 if (newfile != NULL) { 261 newfile = newfile + 1; /* Get rid of leading / */ 262 } else { 263 newfile = file; 264 } 265 266 va_start(adx, fmt); 267 (void) vsnprintf(buf, sizeof (buf), fmt, adx); 268 va_end(adx); 269 270 /* 271 * To get this data, use the zfs-dprintf probe as so: 272 * dtrace -q -n 'zfs-dprintf \ 273 * /stringof(arg0) == "dbuf.c"/ \ 274 * {printf("%s: %s", stringof(arg1), stringof(arg3))}' 275 * arg0 = file name 276 * arg1 = function name 277 * arg2 = line number 278 * arg3 = message 279 */ 280 DTRACE_PROBE4(zfs__dprintf, 281 char *, newfile, char *, func, int, line, char *, buf); 282 } 283 284 static void 285 history_str_free(char *buf) 286 { 287 kmem_free(buf, HIS_MAX_RECORD_LEN); 288 } 289 290 static char * 291 history_str_get(zfs_cmd_t *zc) 292 { 293 char *buf; 294 295 if (zc->zc_history == NULL) 296 return (NULL); 297 298 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP); 299 if (copyinstr((void *)(uintptr_t)zc->zc_history, 300 buf, HIS_MAX_RECORD_LEN, NULL) != 0) { 301 history_str_free(buf); 302 return (NULL); 303 } 304 305 buf[HIS_MAX_RECORD_LEN -1] = '\0'; 306 307 return (buf); 308 } 309 310 /* 311 * Check to see if the named dataset is currently defined as bootable 312 */ 313 static boolean_t 314 zfs_is_bootfs(const char *name) 315 { 316 objset_t *os; 317 318 if (dmu_objset_hold(name, FTAG, &os) == 0) { 319 boolean_t ret; 320 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os))); 321 dmu_objset_rele(os, FTAG); 322 return (ret); 323 } 324 return (B_FALSE); 325 } 326 327 /* 328 * Return non-zero if the spa version is less than requested version. 329 */ 330 static int 331 zfs_earlier_version(const char *name, int version) 332 { 333 spa_t *spa; 334 335 if (spa_open(name, &spa, FTAG) == 0) { 336 if (spa_version(spa) < version) { 337 spa_close(spa, FTAG); 338 return (1); 339 } 340 spa_close(spa, FTAG); 341 } 342 return (0); 343 } 344 345 /* 346 * Return TRUE if the ZPL version is less than requested version. 347 */ 348 static boolean_t 349 zpl_earlier_version(const char *name, int version) 350 { 351 objset_t *os; 352 boolean_t rc = B_TRUE; 353 354 if (dmu_objset_hold(name, FTAG, &os) == 0) { 355 uint64_t zplversion; 356 357 if (dmu_objset_type(os) != DMU_OST_ZFS) { 358 dmu_objset_rele(os, FTAG); 359 return (B_TRUE); 360 } 361 /* XXX reading from non-owned objset */ 362 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0) 363 rc = zplversion < version; 364 dmu_objset_rele(os, FTAG); 365 } 366 return (rc); 367 } 368 369 static void 370 zfs_log_history(zfs_cmd_t *zc) 371 { 372 spa_t *spa; 373 char *buf; 374 375 if ((buf = history_str_get(zc)) == NULL) 376 return; 377 378 if (spa_open(zc->zc_name, &spa, FTAG) == 0) { 379 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY) 380 (void) spa_history_log(spa, buf); 381 spa_close(spa, FTAG); 382 } 383 history_str_free(buf); 384 } 385 386 /* 387 * Policy for top-level read operations (list pools). Requires no privileges, 388 * and can be used in the local zone, as there is no associated dataset. 389 */ 390 /* ARGSUSED */ 391 static int 392 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 393 { 394 return (0); 395 } 396 397 /* 398 * Policy for dataset read operations (list children, get statistics). Requires 399 * no privileges, but must be visible in the local zone. 400 */ 401 /* ARGSUSED */ 402 static int 403 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 404 { 405 if (INGLOBALZONE(curproc) || 406 zone_dataset_visible(zc->zc_name, NULL)) 407 return (0); 408 409 return (SET_ERROR(ENOENT)); 410 } 411 412 static int 413 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr) 414 { 415 int writable = 1; 416 417 /* 418 * The dataset must be visible by this zone -- check this first 419 * so they don't see EPERM on something they shouldn't know about. 420 */ 421 if (!INGLOBALZONE(curproc) && 422 !zone_dataset_visible(dataset, &writable)) 423 return (SET_ERROR(ENOENT)); 424 425 if (INGLOBALZONE(curproc)) { 426 /* 427 * If the fs is zoned, only root can access it from the 428 * global zone. 429 */ 430 if (secpolicy_zfs(cr) && zoned) 431 return (SET_ERROR(EPERM)); 432 } else { 433 /* 434 * If we are in a local zone, the 'zoned' property must be set. 435 */ 436 if (!zoned) 437 return (SET_ERROR(EPERM)); 438 439 /* must be writable by this zone */ 440 if (!writable) 441 return (SET_ERROR(EPERM)); 442 } 443 return (0); 444 } 445 446 static int 447 zfs_dozonecheck(const char *dataset, cred_t *cr) 448 { 449 uint64_t zoned; 450 451 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL)) 452 return (SET_ERROR(ENOENT)); 453 454 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 455 } 456 457 static int 458 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr) 459 { 460 uint64_t zoned; 461 462 if (dsl_prop_get_int_ds(ds, "zoned", &zoned)) 463 return (SET_ERROR(ENOENT)); 464 465 return (zfs_dozonecheck_impl(dataset, zoned, cr)); 466 } 467 468 static int 469 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds, 470 const char *perm, cred_t *cr) 471 { 472 int error; 473 474 error = zfs_dozonecheck_ds(name, ds, cr); 475 if (error == 0) { 476 error = secpolicy_zfs(cr); 477 if (error != 0) 478 error = dsl_deleg_access_impl(ds, perm, cr); 479 } 480 return (error); 481 } 482 483 static int 484 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr) 485 { 486 int error; 487 dsl_dataset_t *ds; 488 dsl_pool_t *dp; 489 490 error = dsl_pool_hold(name, FTAG, &dp); 491 if (error != 0) 492 return (error); 493 494 error = dsl_dataset_hold(dp, name, FTAG, &ds); 495 if (error != 0) { 496 dsl_pool_rele(dp, FTAG); 497 return (error); 498 } 499 500 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr); 501 502 dsl_dataset_rele(ds, FTAG); 503 dsl_pool_rele(dp, FTAG); 504 return (error); 505 } 506 507 /* 508 * Policy for setting the security label property. 509 * 510 * Returns 0 for success, non-zero for access and other errors. 511 */ 512 static int 513 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr) 514 { 515 char ds_hexsl[MAXNAMELEN]; 516 bslabel_t ds_sl, new_sl; 517 boolean_t new_default = FALSE; 518 uint64_t zoned; 519 int needed_priv = -1; 520 int error; 521 522 /* First get the existing dataset label. */ 523 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 524 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 525 if (error != 0) 526 return (SET_ERROR(EPERM)); 527 528 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0) 529 new_default = TRUE; 530 531 /* The label must be translatable */ 532 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0)) 533 return (SET_ERROR(EINVAL)); 534 535 /* 536 * In a non-global zone, disallow attempts to set a label that 537 * doesn't match that of the zone; otherwise no other checks 538 * are needed. 539 */ 540 if (!INGLOBALZONE(curproc)) { 541 if (new_default || !blequal(&new_sl, CR_SL(CRED()))) 542 return (SET_ERROR(EPERM)); 543 return (0); 544 } 545 546 /* 547 * For global-zone datasets (i.e., those whose zoned property is 548 * "off", verify that the specified new label is valid for the 549 * global zone. 550 */ 551 if (dsl_prop_get_integer(name, 552 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 553 return (SET_ERROR(EPERM)); 554 if (!zoned) { 555 if (zfs_check_global_label(name, strval) != 0) 556 return (SET_ERROR(EPERM)); 557 } 558 559 /* 560 * If the existing dataset label is nondefault, check if the 561 * dataset is mounted (label cannot be changed while mounted). 562 * Get the zfsvfs; if there isn't one, then the dataset isn't 563 * mounted (or isn't a dataset, doesn't exist, ...). 564 */ 565 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) { 566 objset_t *os; 567 static char *setsl_tag = "setsl_tag"; 568 569 /* 570 * Try to own the dataset; abort if there is any error, 571 * (e.g., already mounted, in use, or other error). 572 */ 573 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, 574 setsl_tag, &os); 575 if (error != 0) 576 return (SET_ERROR(EPERM)); 577 578 dmu_objset_disown(os, setsl_tag); 579 580 if (new_default) { 581 needed_priv = PRIV_FILE_DOWNGRADE_SL; 582 goto out_check; 583 } 584 585 if (hexstr_to_label(strval, &new_sl) != 0) 586 return (SET_ERROR(EPERM)); 587 588 if (blstrictdom(&ds_sl, &new_sl)) 589 needed_priv = PRIV_FILE_DOWNGRADE_SL; 590 else if (blstrictdom(&new_sl, &ds_sl)) 591 needed_priv = PRIV_FILE_UPGRADE_SL; 592 } else { 593 /* dataset currently has a default label */ 594 if (!new_default) 595 needed_priv = PRIV_FILE_UPGRADE_SL; 596 } 597 598 out_check: 599 if (needed_priv != -1) 600 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL)); 601 return (0); 602 } 603 604 static int 605 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval, 606 cred_t *cr) 607 { 608 char *strval; 609 610 /* 611 * Check permissions for special properties. 612 */ 613 switch (prop) { 614 case ZFS_PROP_ZONED: 615 /* 616 * Disallow setting of 'zoned' from within a local zone. 617 */ 618 if (!INGLOBALZONE(curproc)) 619 return (SET_ERROR(EPERM)); 620 break; 621 622 case ZFS_PROP_QUOTA: 623 case ZFS_PROP_FILESYSTEM_LIMIT: 624 case ZFS_PROP_SNAPSHOT_LIMIT: 625 if (!INGLOBALZONE(curproc)) { 626 uint64_t zoned; 627 char setpoint[MAXNAMELEN]; 628 /* 629 * Unprivileged users are allowed to modify the 630 * limit on things *under* (ie. contained by) 631 * the thing they own. 632 */ 633 if (dsl_prop_get_integer(dsname, "zoned", &zoned, 634 setpoint)) 635 return (SET_ERROR(EPERM)); 636 if (!zoned || strlen(dsname) <= strlen(setpoint)) 637 return (SET_ERROR(EPERM)); 638 } 639 break; 640 641 case ZFS_PROP_MLSLABEL: 642 if (!is_system_labeled()) 643 return (SET_ERROR(EPERM)); 644 645 if (nvpair_value_string(propval, &strval) == 0) { 646 int err; 647 648 err = zfs_set_slabel_policy(dsname, strval, CRED()); 649 if (err != 0) 650 return (err); 651 } 652 break; 653 } 654 655 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr)); 656 } 657 658 /* ARGSUSED */ 659 static int 660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 661 { 662 int error; 663 664 error = zfs_dozonecheck(zc->zc_name, cr); 665 if (error != 0) 666 return (error); 667 668 /* 669 * permission to set permissions will be evaluated later in 670 * dsl_deleg_can_allow() 671 */ 672 return (0); 673 } 674 675 /* ARGSUSED */ 676 static int 677 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 678 { 679 return (zfs_secpolicy_write_perms(zc->zc_name, 680 ZFS_DELEG_PERM_ROLLBACK, cr)); 681 } 682 683 /* ARGSUSED */ 684 static int 685 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 686 { 687 dsl_pool_t *dp; 688 dsl_dataset_t *ds; 689 char *cp; 690 int error; 691 692 /* 693 * Generate the current snapshot name from the given objsetid, then 694 * use that name for the secpolicy/zone checks. 695 */ 696 cp = strchr(zc->zc_name, '@'); 697 if (cp == NULL) 698 return (SET_ERROR(EINVAL)); 699 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 700 if (error != 0) 701 return (error); 702 703 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds); 704 if (error != 0) { 705 dsl_pool_rele(dp, FTAG); 706 return (error); 707 } 708 709 dsl_dataset_name(ds, zc->zc_name); 710 711 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds, 712 ZFS_DELEG_PERM_SEND, cr); 713 dsl_dataset_rele(ds, FTAG); 714 dsl_pool_rele(dp, FTAG); 715 716 return (error); 717 } 718 719 /* ARGSUSED */ 720 static int 721 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 722 { 723 return (zfs_secpolicy_write_perms(zc->zc_name, 724 ZFS_DELEG_PERM_SEND, cr)); 725 } 726 727 /* ARGSUSED */ 728 static int 729 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 730 { 731 vnode_t *vp; 732 int error; 733 734 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 735 NO_FOLLOW, NULL, &vp)) != 0) 736 return (error); 737 738 /* Now make sure mntpnt and dataset are ZFS */ 739 740 if (vp->v_vfsp->vfs_fstype != zfsfstype || 741 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 742 zc->zc_name) != 0)) { 743 VN_RELE(vp); 744 return (SET_ERROR(EPERM)); 745 } 746 747 VN_RELE(vp); 748 return (dsl_deleg_access(zc->zc_name, 749 ZFS_DELEG_PERM_SHARE, cr)); 750 } 751 752 int 753 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 754 { 755 if (!INGLOBALZONE(curproc)) 756 return (SET_ERROR(EPERM)); 757 758 if (secpolicy_nfs(cr) == 0) { 759 return (0); 760 } else { 761 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 762 } 763 } 764 765 int 766 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 767 { 768 if (!INGLOBALZONE(curproc)) 769 return (SET_ERROR(EPERM)); 770 771 if (secpolicy_smb(cr) == 0) { 772 return (0); 773 } else { 774 return (zfs_secpolicy_deleg_share(zc, innvl, cr)); 775 } 776 } 777 778 static int 779 zfs_get_parent(const char *datasetname, char *parent, int parentsize) 780 { 781 char *cp; 782 783 /* 784 * Remove the @bla or /bla from the end of the name to get the parent. 785 */ 786 (void) strncpy(parent, datasetname, parentsize); 787 cp = strrchr(parent, '@'); 788 if (cp != NULL) { 789 cp[0] = '\0'; 790 } else { 791 cp = strrchr(parent, '/'); 792 if (cp == NULL) 793 return (SET_ERROR(ENOENT)); 794 cp[0] = '\0'; 795 } 796 797 return (0); 798 } 799 800 int 801 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 802 { 803 int error; 804 805 if ((error = zfs_secpolicy_write_perms(name, 806 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 807 return (error); 808 809 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr)); 810 } 811 812 /* ARGSUSED */ 813 static int 814 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 815 { 816 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr)); 817 } 818 819 /* 820 * Destroying snapshots with delegated permissions requires 821 * descendant mount and destroy permissions. 822 */ 823 /* ARGSUSED */ 824 static int 825 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 826 { 827 nvlist_t *snaps; 828 nvpair_t *pair, *nextpair; 829 int error = 0; 830 831 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 832 return (SET_ERROR(EINVAL)); 833 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 834 pair = nextpair) { 835 nextpair = nvlist_next_nvpair(snaps, pair); 836 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr); 837 if (error == ENOENT) { 838 /* 839 * Ignore any snapshots that don't exist (we consider 840 * them "already destroyed"). Remove the name from the 841 * nvl here in case the snapshot is created between 842 * now and when we try to destroy it (in which case 843 * we don't want to destroy it since we haven't 844 * checked for permission). 845 */ 846 fnvlist_remove_nvpair(snaps, pair); 847 error = 0; 848 } 849 if (error != 0) 850 break; 851 } 852 853 return (error); 854 } 855 856 int 857 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 858 { 859 char parentname[MAXNAMELEN]; 860 int error; 861 862 if ((error = zfs_secpolicy_write_perms(from, 863 ZFS_DELEG_PERM_RENAME, cr)) != 0) 864 return (error); 865 866 if ((error = zfs_secpolicy_write_perms(from, 867 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 868 return (error); 869 870 if ((error = zfs_get_parent(to, parentname, 871 sizeof (parentname))) != 0) 872 return (error); 873 874 if ((error = zfs_secpolicy_write_perms(parentname, 875 ZFS_DELEG_PERM_CREATE, cr)) != 0) 876 return (error); 877 878 if ((error = zfs_secpolicy_write_perms(parentname, 879 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 880 return (error); 881 882 return (error); 883 } 884 885 /* ARGSUSED */ 886 static int 887 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 888 { 889 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr)); 890 } 891 892 /* ARGSUSED */ 893 static int 894 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 895 { 896 dsl_pool_t *dp; 897 dsl_dataset_t *clone; 898 int error; 899 900 error = zfs_secpolicy_write_perms(zc->zc_name, 901 ZFS_DELEG_PERM_PROMOTE, cr); 902 if (error != 0) 903 return (error); 904 905 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 906 if (error != 0) 907 return (error); 908 909 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone); 910 911 if (error == 0) { 912 char parentname[MAXNAMELEN]; 913 dsl_dataset_t *origin = NULL; 914 dsl_dir_t *dd; 915 dd = clone->ds_dir; 916 917 error = dsl_dataset_hold_obj(dd->dd_pool, 918 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin); 919 if (error != 0) { 920 dsl_dataset_rele(clone, FTAG); 921 dsl_pool_rele(dp, FTAG); 922 return (error); 923 } 924 925 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone, 926 ZFS_DELEG_PERM_MOUNT, cr); 927 928 dsl_dataset_name(origin, parentname); 929 if (error == 0) { 930 error = zfs_secpolicy_write_perms_ds(parentname, origin, 931 ZFS_DELEG_PERM_PROMOTE, cr); 932 } 933 dsl_dataset_rele(clone, FTAG); 934 dsl_dataset_rele(origin, FTAG); 935 } 936 dsl_pool_rele(dp, FTAG); 937 return (error); 938 } 939 940 /* ARGSUSED */ 941 static int 942 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 943 { 944 int error; 945 946 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 947 ZFS_DELEG_PERM_RECEIVE, cr)) != 0) 948 return (error); 949 950 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 951 ZFS_DELEG_PERM_MOUNT, cr)) != 0) 952 return (error); 953 954 return (zfs_secpolicy_write_perms(zc->zc_name, 955 ZFS_DELEG_PERM_CREATE, cr)); 956 } 957 958 int 959 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 960 { 961 return (zfs_secpolicy_write_perms(name, 962 ZFS_DELEG_PERM_SNAPSHOT, cr)); 963 } 964 965 /* 966 * Check for permission to create each snapshot in the nvlist. 967 */ 968 /* ARGSUSED */ 969 static int 970 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 971 { 972 nvlist_t *snaps; 973 int error = 0; 974 nvpair_t *pair; 975 976 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 977 return (SET_ERROR(EINVAL)); 978 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 979 pair = nvlist_next_nvpair(snaps, pair)) { 980 char *name = nvpair_name(pair); 981 char *atp = strchr(name, '@'); 982 983 if (atp == NULL) { 984 error = SET_ERROR(EINVAL); 985 break; 986 } 987 *atp = '\0'; 988 error = zfs_secpolicy_snapshot_perms(name, cr); 989 *atp = '@'; 990 if (error != 0) 991 break; 992 } 993 return (error); 994 } 995 996 /* 997 * Check for permission to create each snapshot in the nvlist. 998 */ 999 /* ARGSUSED */ 1000 static int 1001 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1002 { 1003 int error = 0; 1004 1005 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 1006 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 1007 char *name = nvpair_name(pair); 1008 char *hashp = strchr(name, '#'); 1009 1010 if (hashp == NULL) { 1011 error = SET_ERROR(EINVAL); 1012 break; 1013 } 1014 *hashp = '\0'; 1015 error = zfs_secpolicy_write_perms(name, 1016 ZFS_DELEG_PERM_BOOKMARK, cr); 1017 *hashp = '#'; 1018 if (error != 0) 1019 break; 1020 } 1021 return (error); 1022 } 1023 1024 /* ARGSUSED */ 1025 static int 1026 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1027 { 1028 nvpair_t *pair, *nextpair; 1029 int error = 0; 1030 1031 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1032 pair = nextpair) { 1033 char *name = nvpair_name(pair); 1034 char *hashp = strchr(name, '#'); 1035 nextpair = nvlist_next_nvpair(innvl, pair); 1036 1037 if (hashp == NULL) { 1038 error = SET_ERROR(EINVAL); 1039 break; 1040 } 1041 1042 *hashp = '\0'; 1043 error = zfs_secpolicy_write_perms(name, 1044 ZFS_DELEG_PERM_DESTROY, cr); 1045 *hashp = '#'; 1046 if (error == ENOENT) { 1047 /* 1048 * Ignore any filesystems that don't exist (we consider 1049 * their bookmarks "already destroyed"). Remove 1050 * the name from the nvl here in case the filesystem 1051 * is created between now and when we try to destroy 1052 * the bookmark (in which case we don't want to 1053 * destroy it since we haven't checked for permission). 1054 */ 1055 fnvlist_remove_nvpair(innvl, pair); 1056 error = 0; 1057 } 1058 if (error != 0) 1059 break; 1060 } 1061 1062 return (error); 1063 } 1064 1065 /* ARGSUSED */ 1066 static int 1067 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1068 { 1069 /* 1070 * Even root must have a proper TSD so that we know what pool 1071 * to log to. 1072 */ 1073 if (tsd_get(zfs_allow_log_key) == NULL) 1074 return (SET_ERROR(EPERM)); 1075 return (0); 1076 } 1077 1078 static int 1079 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1080 { 1081 char parentname[MAXNAMELEN]; 1082 int error; 1083 char *origin; 1084 1085 if ((error = zfs_get_parent(zc->zc_name, parentname, 1086 sizeof (parentname))) != 0) 1087 return (error); 1088 1089 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 && 1090 (error = zfs_secpolicy_write_perms(origin, 1091 ZFS_DELEG_PERM_CLONE, cr)) != 0) 1092 return (error); 1093 1094 if ((error = zfs_secpolicy_write_perms(parentname, 1095 ZFS_DELEG_PERM_CREATE, cr)) != 0) 1096 return (error); 1097 1098 return (zfs_secpolicy_write_perms(parentname, 1099 ZFS_DELEG_PERM_MOUNT, cr)); 1100 } 1101 1102 /* 1103 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires 1104 * SYS_CONFIG privilege, which is not available in a local zone. 1105 */ 1106 /* ARGSUSED */ 1107 static int 1108 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1109 { 1110 if (secpolicy_sys_config(cr, B_FALSE) != 0) 1111 return (SET_ERROR(EPERM)); 1112 1113 return (0); 1114 } 1115 1116 /* 1117 * Policy for object to name lookups. 1118 */ 1119 /* ARGSUSED */ 1120 static int 1121 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1122 { 1123 int error; 1124 1125 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0) 1126 return (0); 1127 1128 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr); 1129 return (error); 1130 } 1131 1132 /* 1133 * Policy for fault injection. Requires all privileges. 1134 */ 1135 /* ARGSUSED */ 1136 static int 1137 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1138 { 1139 return (secpolicy_zinject(cr)); 1140 } 1141 1142 /* ARGSUSED */ 1143 static int 1144 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1145 { 1146 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value); 1147 1148 if (prop == ZPROP_INVAL) { 1149 if (!zfs_prop_user(zc->zc_value)) 1150 return (SET_ERROR(EINVAL)); 1151 return (zfs_secpolicy_write_perms(zc->zc_name, 1152 ZFS_DELEG_PERM_USERPROP, cr)); 1153 } else { 1154 return (zfs_secpolicy_setprop(zc->zc_name, prop, 1155 NULL, cr)); 1156 } 1157 } 1158 1159 static int 1160 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1161 { 1162 int err = zfs_secpolicy_read(zc, innvl, cr); 1163 if (err) 1164 return (err); 1165 1166 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1167 return (SET_ERROR(EINVAL)); 1168 1169 if (zc->zc_value[0] == 0) { 1170 /* 1171 * They are asking about a posix uid/gid. If it's 1172 * themself, allow it. 1173 */ 1174 if (zc->zc_objset_type == ZFS_PROP_USERUSED || 1175 zc->zc_objset_type == ZFS_PROP_USERQUOTA) { 1176 if (zc->zc_guid == crgetuid(cr)) 1177 return (0); 1178 } else { 1179 if (groupmember(zc->zc_guid, cr)) 1180 return (0); 1181 } 1182 } 1183 1184 return (zfs_secpolicy_write_perms(zc->zc_name, 1185 userquota_perms[zc->zc_objset_type], cr)); 1186 } 1187 1188 static int 1189 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1190 { 1191 int err = zfs_secpolicy_read(zc, innvl, cr); 1192 if (err) 1193 return (err); 1194 1195 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 1196 return (SET_ERROR(EINVAL)); 1197 1198 return (zfs_secpolicy_write_perms(zc->zc_name, 1199 userquota_perms[zc->zc_objset_type], cr)); 1200 } 1201 1202 /* ARGSUSED */ 1203 static int 1204 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1205 { 1206 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, 1207 NULL, cr)); 1208 } 1209 1210 /* ARGSUSED */ 1211 static int 1212 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1213 { 1214 nvpair_t *pair; 1215 nvlist_t *holds; 1216 int error; 1217 1218 error = nvlist_lookup_nvlist(innvl, "holds", &holds); 1219 if (error != 0) 1220 return (SET_ERROR(EINVAL)); 1221 1222 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 1223 pair = nvlist_next_nvpair(holds, pair)) { 1224 char fsname[MAXNAMELEN]; 1225 error = dmu_fsname(nvpair_name(pair), fsname); 1226 if (error != 0) 1227 return (error); 1228 error = zfs_secpolicy_write_perms(fsname, 1229 ZFS_DELEG_PERM_HOLD, cr); 1230 if (error != 0) 1231 return (error); 1232 } 1233 return (0); 1234 } 1235 1236 /* ARGSUSED */ 1237 static int 1238 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1239 { 1240 nvpair_t *pair; 1241 int error; 1242 1243 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL; 1244 pair = nvlist_next_nvpair(innvl, pair)) { 1245 char fsname[MAXNAMELEN]; 1246 error = dmu_fsname(nvpair_name(pair), fsname); 1247 if (error != 0) 1248 return (error); 1249 error = zfs_secpolicy_write_perms(fsname, 1250 ZFS_DELEG_PERM_RELEASE, cr); 1251 if (error != 0) 1252 return (error); 1253 } 1254 return (0); 1255 } 1256 1257 /* 1258 * Policy for allowing temporary snapshots to be taken or released 1259 */ 1260 static int 1261 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr) 1262 { 1263 /* 1264 * A temporary snapshot is the same as a snapshot, 1265 * hold, destroy and release all rolled into one. 1266 * Delegated diff alone is sufficient that we allow this. 1267 */ 1268 int error; 1269 1270 if ((error = zfs_secpolicy_write_perms(zc->zc_name, 1271 ZFS_DELEG_PERM_DIFF, cr)) == 0) 1272 return (0); 1273 1274 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr); 1275 if (error == 0) 1276 error = zfs_secpolicy_hold(zc, innvl, cr); 1277 if (error == 0) 1278 error = zfs_secpolicy_release(zc, innvl, cr); 1279 if (error == 0) 1280 error = zfs_secpolicy_destroy(zc, innvl, cr); 1281 return (error); 1282 } 1283 1284 /* 1285 * Returns the nvlist as specified by the user in the zfs_cmd_t. 1286 */ 1287 static int 1288 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp) 1289 { 1290 char *packed; 1291 int error; 1292 nvlist_t *list = NULL; 1293 1294 /* 1295 * Read in and unpack the user-supplied nvlist. 1296 */ 1297 if (size == 0) 1298 return (SET_ERROR(EINVAL)); 1299 1300 packed = kmem_alloc(size, KM_SLEEP); 1301 1302 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size, 1303 iflag)) != 0) { 1304 kmem_free(packed, size); 1305 return (error); 1306 } 1307 1308 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) { 1309 kmem_free(packed, size); 1310 return (error); 1311 } 1312 1313 kmem_free(packed, size); 1314 1315 *nvp = list; 1316 return (0); 1317 } 1318 1319 /* 1320 * Reduce the size of this nvlist until it can be serialized in 'max' bytes. 1321 * Entries will be removed from the end of the nvlist, and one int32 entry 1322 * named "N_MORE_ERRORS" will be added indicating how many entries were 1323 * removed. 1324 */ 1325 static int 1326 nvlist_smush(nvlist_t *errors, size_t max) 1327 { 1328 size_t size; 1329 1330 size = fnvlist_size(errors); 1331 1332 if (size > max) { 1333 nvpair_t *more_errors; 1334 int n = 0; 1335 1336 if (max < 1024) 1337 return (SET_ERROR(ENOMEM)); 1338 1339 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0); 1340 more_errors = nvlist_prev_nvpair(errors, NULL); 1341 1342 do { 1343 nvpair_t *pair = nvlist_prev_nvpair(errors, 1344 more_errors); 1345 fnvlist_remove_nvpair(errors, pair); 1346 n++; 1347 size = fnvlist_size(errors); 1348 } while (size > max); 1349 1350 fnvlist_remove_nvpair(errors, more_errors); 1351 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n); 1352 ASSERT3U(fnvlist_size(errors), <=, max); 1353 } 1354 1355 return (0); 1356 } 1357 1358 static int 1359 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl) 1360 { 1361 char *packed = NULL; 1362 int error = 0; 1363 size_t size; 1364 1365 size = fnvlist_size(nvl); 1366 1367 if (size > zc->zc_nvlist_dst_size) { 1368 error = SET_ERROR(ENOMEM); 1369 } else { 1370 packed = fnvlist_pack(nvl, &size); 1371 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst, 1372 size, zc->zc_iflags) != 0) 1373 error = SET_ERROR(EFAULT); 1374 fnvlist_pack_free(packed, size); 1375 } 1376 1377 zc->zc_nvlist_dst_size = size; 1378 zc->zc_nvlist_dst_filled = B_TRUE; 1379 return (error); 1380 } 1381 1382 static int 1383 getzfsvfs(const char *dsname, zfsvfs_t **zfvp) 1384 { 1385 objset_t *os; 1386 int error; 1387 1388 error = dmu_objset_hold(dsname, FTAG, &os); 1389 if (error != 0) 1390 return (error); 1391 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1392 dmu_objset_rele(os, FTAG); 1393 return (SET_ERROR(EINVAL)); 1394 } 1395 1396 mutex_enter(&os->os_user_ptr_lock); 1397 *zfvp = dmu_objset_get_user(os); 1398 if (*zfvp) { 1399 VFS_HOLD((*zfvp)->z_vfs); 1400 } else { 1401 error = SET_ERROR(ESRCH); 1402 } 1403 mutex_exit(&os->os_user_ptr_lock); 1404 dmu_objset_rele(os, FTAG); 1405 return (error); 1406 } 1407 1408 /* 1409 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which 1410 * case its z_vfs will be NULL, and it will be opened as the owner. 1411 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER, 1412 * which prevents all vnode ops from running. 1413 */ 1414 static int 1415 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer) 1416 { 1417 int error = 0; 1418 1419 if (getzfsvfs(name, zfvp) != 0) 1420 error = zfsvfs_create(name, zfvp); 1421 if (error == 0) { 1422 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER : 1423 RW_READER, tag); 1424 if ((*zfvp)->z_unmounted) { 1425 /* 1426 * XXX we could probably try again, since the unmounting 1427 * thread should be just about to disassociate the 1428 * objset from the zfsvfs. 1429 */ 1430 rrm_exit(&(*zfvp)->z_teardown_lock, tag); 1431 return (SET_ERROR(EBUSY)); 1432 } 1433 } 1434 return (error); 1435 } 1436 1437 static void 1438 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag) 1439 { 1440 rrm_exit(&zfsvfs->z_teardown_lock, tag); 1441 1442 if (zfsvfs->z_vfs) { 1443 VFS_RELE(zfsvfs->z_vfs); 1444 } else { 1445 dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1446 zfsvfs_free(zfsvfs); 1447 } 1448 } 1449 1450 static int 1451 zfs_ioc_pool_create(zfs_cmd_t *zc) 1452 { 1453 int error; 1454 nvlist_t *config, *props = NULL; 1455 nvlist_t *rootprops = NULL; 1456 nvlist_t *zplprops = NULL; 1457 1458 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1459 zc->zc_iflags, &config)) 1460 return (error); 1461 1462 if (zc->zc_nvlist_src_size != 0 && (error = 1463 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1464 zc->zc_iflags, &props))) { 1465 nvlist_free(config); 1466 return (error); 1467 } 1468 1469 if (props) { 1470 nvlist_t *nvl = NULL; 1471 uint64_t version = SPA_VERSION; 1472 1473 (void) nvlist_lookup_uint64(props, 1474 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version); 1475 if (!SPA_VERSION_IS_SUPPORTED(version)) { 1476 error = SET_ERROR(EINVAL); 1477 goto pool_props_bad; 1478 } 1479 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl); 1480 if (nvl) { 1481 error = nvlist_dup(nvl, &rootprops, KM_SLEEP); 1482 if (error != 0) { 1483 nvlist_free(config); 1484 nvlist_free(props); 1485 return (error); 1486 } 1487 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS); 1488 } 1489 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1490 error = zfs_fill_zplprops_root(version, rootprops, 1491 zplprops, NULL); 1492 if (error != 0) 1493 goto pool_props_bad; 1494 } 1495 1496 error = spa_create(zc->zc_name, config, props, zplprops); 1497 1498 /* 1499 * Set the remaining root properties 1500 */ 1501 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name, 1502 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) 1503 (void) spa_destroy(zc->zc_name); 1504 1505 pool_props_bad: 1506 nvlist_free(rootprops); 1507 nvlist_free(zplprops); 1508 nvlist_free(config); 1509 nvlist_free(props); 1510 1511 return (error); 1512 } 1513 1514 static int 1515 zfs_ioc_pool_destroy(zfs_cmd_t *zc) 1516 { 1517 int error; 1518 zfs_log_history(zc); 1519 error = spa_destroy(zc->zc_name); 1520 if (error == 0) 1521 zvol_remove_minors(zc->zc_name); 1522 return (error); 1523 } 1524 1525 static int 1526 zfs_ioc_pool_import(zfs_cmd_t *zc) 1527 { 1528 nvlist_t *config, *props = NULL; 1529 uint64_t guid; 1530 int error; 1531 1532 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1533 zc->zc_iflags, &config)) != 0) 1534 return (error); 1535 1536 if (zc->zc_nvlist_src_size != 0 && (error = 1537 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1538 zc->zc_iflags, &props))) { 1539 nvlist_free(config); 1540 return (error); 1541 } 1542 1543 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 || 1544 guid != zc->zc_guid) 1545 error = SET_ERROR(EINVAL); 1546 else 1547 error = spa_import(zc->zc_name, config, props, zc->zc_cookie); 1548 1549 if (zc->zc_nvlist_dst != 0) { 1550 int err; 1551 1552 if ((err = put_nvlist(zc, config)) != 0) 1553 error = err; 1554 } 1555 1556 nvlist_free(config); 1557 1558 if (props) 1559 nvlist_free(props); 1560 1561 return (error); 1562 } 1563 1564 static int 1565 zfs_ioc_pool_export(zfs_cmd_t *zc) 1566 { 1567 int error; 1568 boolean_t force = (boolean_t)zc->zc_cookie; 1569 boolean_t hardforce = (boolean_t)zc->zc_guid; 1570 1571 zfs_log_history(zc); 1572 error = spa_export(zc->zc_name, NULL, force, hardforce); 1573 if (error == 0) 1574 zvol_remove_minors(zc->zc_name); 1575 return (error); 1576 } 1577 1578 static int 1579 zfs_ioc_pool_configs(zfs_cmd_t *zc) 1580 { 1581 nvlist_t *configs; 1582 int error; 1583 1584 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL) 1585 return (SET_ERROR(EEXIST)); 1586 1587 error = put_nvlist(zc, configs); 1588 1589 nvlist_free(configs); 1590 1591 return (error); 1592 } 1593 1594 /* 1595 * inputs: 1596 * zc_name name of the pool 1597 * 1598 * outputs: 1599 * zc_cookie real errno 1600 * zc_nvlist_dst config nvlist 1601 * zc_nvlist_dst_size size of config nvlist 1602 */ 1603 static int 1604 zfs_ioc_pool_stats(zfs_cmd_t *zc) 1605 { 1606 nvlist_t *config; 1607 int error; 1608 int ret = 0; 1609 1610 error = spa_get_stats(zc->zc_name, &config, zc->zc_value, 1611 sizeof (zc->zc_value)); 1612 1613 if (config != NULL) { 1614 ret = put_nvlist(zc, config); 1615 nvlist_free(config); 1616 1617 /* 1618 * The config may be present even if 'error' is non-zero. 1619 * In this case we return success, and preserve the real errno 1620 * in 'zc_cookie'. 1621 */ 1622 zc->zc_cookie = error; 1623 } else { 1624 ret = error; 1625 } 1626 1627 return (ret); 1628 } 1629 1630 /* 1631 * Try to import the given pool, returning pool stats as appropriate so that 1632 * user land knows which devices are available and overall pool health. 1633 */ 1634 static int 1635 zfs_ioc_pool_tryimport(zfs_cmd_t *zc) 1636 { 1637 nvlist_t *tryconfig, *config; 1638 int error; 1639 1640 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1641 zc->zc_iflags, &tryconfig)) != 0) 1642 return (error); 1643 1644 config = spa_tryimport(tryconfig); 1645 1646 nvlist_free(tryconfig); 1647 1648 if (config == NULL) 1649 return (SET_ERROR(EINVAL)); 1650 1651 error = put_nvlist(zc, config); 1652 nvlist_free(config); 1653 1654 return (error); 1655 } 1656 1657 /* 1658 * inputs: 1659 * zc_name name of the pool 1660 * zc_cookie scan func (pool_scan_func_t) 1661 */ 1662 static int 1663 zfs_ioc_pool_scan(zfs_cmd_t *zc) 1664 { 1665 spa_t *spa; 1666 int error; 1667 1668 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1669 return (error); 1670 1671 if (zc->zc_cookie == POOL_SCAN_NONE) 1672 error = spa_scan_stop(spa); 1673 else 1674 error = spa_scan(spa, zc->zc_cookie); 1675 1676 spa_close(spa, FTAG); 1677 1678 return (error); 1679 } 1680 1681 static int 1682 zfs_ioc_pool_freeze(zfs_cmd_t *zc) 1683 { 1684 spa_t *spa; 1685 int error; 1686 1687 error = spa_open(zc->zc_name, &spa, FTAG); 1688 if (error == 0) { 1689 spa_freeze(spa); 1690 spa_close(spa, FTAG); 1691 } 1692 return (error); 1693 } 1694 1695 static int 1696 zfs_ioc_pool_upgrade(zfs_cmd_t *zc) 1697 { 1698 spa_t *spa; 1699 int error; 1700 1701 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1702 return (error); 1703 1704 if (zc->zc_cookie < spa_version(spa) || 1705 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) { 1706 spa_close(spa, FTAG); 1707 return (SET_ERROR(EINVAL)); 1708 } 1709 1710 spa_upgrade(spa, zc->zc_cookie); 1711 spa_close(spa, FTAG); 1712 1713 return (error); 1714 } 1715 1716 static int 1717 zfs_ioc_pool_get_history(zfs_cmd_t *zc) 1718 { 1719 spa_t *spa; 1720 char *hist_buf; 1721 uint64_t size; 1722 int error; 1723 1724 if ((size = zc->zc_history_len) == 0) 1725 return (SET_ERROR(EINVAL)); 1726 1727 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1728 return (error); 1729 1730 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 1731 spa_close(spa, FTAG); 1732 return (SET_ERROR(ENOTSUP)); 1733 } 1734 1735 hist_buf = kmem_alloc(size, KM_SLEEP); 1736 if ((error = spa_history_get(spa, &zc->zc_history_offset, 1737 &zc->zc_history_len, hist_buf)) == 0) { 1738 error = ddi_copyout(hist_buf, 1739 (void *)(uintptr_t)zc->zc_history, 1740 zc->zc_history_len, zc->zc_iflags); 1741 } 1742 1743 spa_close(spa, FTAG); 1744 kmem_free(hist_buf, size); 1745 return (error); 1746 } 1747 1748 static int 1749 zfs_ioc_pool_reguid(zfs_cmd_t *zc) 1750 { 1751 spa_t *spa; 1752 int error; 1753 1754 error = spa_open(zc->zc_name, &spa, FTAG); 1755 if (error == 0) { 1756 error = spa_change_guid(spa); 1757 spa_close(spa, FTAG); 1758 } 1759 return (error); 1760 } 1761 1762 static int 1763 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc) 1764 { 1765 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value)); 1766 } 1767 1768 /* 1769 * inputs: 1770 * zc_name name of filesystem 1771 * zc_obj object to find 1772 * 1773 * outputs: 1774 * zc_value name of object 1775 */ 1776 static int 1777 zfs_ioc_obj_to_path(zfs_cmd_t *zc) 1778 { 1779 objset_t *os; 1780 int error; 1781 1782 /* XXX reading from objset not owned */ 1783 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1784 return (error); 1785 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1786 dmu_objset_rele(os, FTAG); 1787 return (SET_ERROR(EINVAL)); 1788 } 1789 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value, 1790 sizeof (zc->zc_value)); 1791 dmu_objset_rele(os, FTAG); 1792 1793 return (error); 1794 } 1795 1796 /* 1797 * inputs: 1798 * zc_name name of filesystem 1799 * zc_obj object to find 1800 * 1801 * outputs: 1802 * zc_stat stats on object 1803 * zc_value path to object 1804 */ 1805 static int 1806 zfs_ioc_obj_to_stats(zfs_cmd_t *zc) 1807 { 1808 objset_t *os; 1809 int error; 1810 1811 /* XXX reading from objset not owned */ 1812 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0) 1813 return (error); 1814 if (dmu_objset_type(os) != DMU_OST_ZFS) { 1815 dmu_objset_rele(os, FTAG); 1816 return (SET_ERROR(EINVAL)); 1817 } 1818 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value, 1819 sizeof (zc->zc_value)); 1820 dmu_objset_rele(os, FTAG); 1821 1822 return (error); 1823 } 1824 1825 static int 1826 zfs_ioc_vdev_add(zfs_cmd_t *zc) 1827 { 1828 spa_t *spa; 1829 int error; 1830 nvlist_t *config, **l2cache, **spares; 1831 uint_t nl2cache = 0, nspares = 0; 1832 1833 error = spa_open(zc->zc_name, &spa, FTAG); 1834 if (error != 0) 1835 return (error); 1836 1837 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1838 zc->zc_iflags, &config); 1839 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE, 1840 &l2cache, &nl2cache); 1841 1842 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES, 1843 &spares, &nspares); 1844 1845 /* 1846 * A root pool with concatenated devices is not supported. 1847 * Thus, can not add a device to a root pool. 1848 * 1849 * Intent log device can not be added to a rootpool because 1850 * during mountroot, zil is replayed, a seperated log device 1851 * can not be accessed during the mountroot time. 1852 * 1853 * l2cache and spare devices are ok to be added to a rootpool. 1854 */ 1855 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) { 1856 nvlist_free(config); 1857 spa_close(spa, FTAG); 1858 return (SET_ERROR(EDOM)); 1859 } 1860 1861 if (error == 0) { 1862 error = spa_vdev_add(spa, config); 1863 nvlist_free(config); 1864 } 1865 spa_close(spa, FTAG); 1866 return (error); 1867 } 1868 1869 /* 1870 * inputs: 1871 * zc_name name of the pool 1872 * zc_nvlist_conf nvlist of devices to remove 1873 * zc_cookie to stop the remove? 1874 */ 1875 static int 1876 zfs_ioc_vdev_remove(zfs_cmd_t *zc) 1877 { 1878 spa_t *spa; 1879 int error; 1880 1881 error = spa_open(zc->zc_name, &spa, FTAG); 1882 if (error != 0) 1883 return (error); 1884 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE); 1885 spa_close(spa, FTAG); 1886 return (error); 1887 } 1888 1889 static int 1890 zfs_ioc_vdev_set_state(zfs_cmd_t *zc) 1891 { 1892 spa_t *spa; 1893 int error; 1894 vdev_state_t newstate = VDEV_STATE_UNKNOWN; 1895 1896 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1897 return (error); 1898 switch (zc->zc_cookie) { 1899 case VDEV_STATE_ONLINE: 1900 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate); 1901 break; 1902 1903 case VDEV_STATE_OFFLINE: 1904 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj); 1905 break; 1906 1907 case VDEV_STATE_FAULTED: 1908 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1909 zc->zc_obj != VDEV_AUX_EXTERNAL) 1910 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1911 1912 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj); 1913 break; 1914 1915 case VDEV_STATE_DEGRADED: 1916 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED && 1917 zc->zc_obj != VDEV_AUX_EXTERNAL) 1918 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED; 1919 1920 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj); 1921 break; 1922 1923 default: 1924 error = SET_ERROR(EINVAL); 1925 } 1926 zc->zc_cookie = newstate; 1927 spa_close(spa, FTAG); 1928 return (error); 1929 } 1930 1931 static int 1932 zfs_ioc_vdev_attach(zfs_cmd_t *zc) 1933 { 1934 spa_t *spa; 1935 int replacing = zc->zc_cookie; 1936 nvlist_t *config; 1937 int error; 1938 1939 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1940 return (error); 1941 1942 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1943 zc->zc_iflags, &config)) == 0) { 1944 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing); 1945 nvlist_free(config); 1946 } 1947 1948 spa_close(spa, FTAG); 1949 return (error); 1950 } 1951 1952 static int 1953 zfs_ioc_vdev_detach(zfs_cmd_t *zc) 1954 { 1955 spa_t *spa; 1956 int error; 1957 1958 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1959 return (error); 1960 1961 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE); 1962 1963 spa_close(spa, FTAG); 1964 return (error); 1965 } 1966 1967 static int 1968 zfs_ioc_vdev_split(zfs_cmd_t *zc) 1969 { 1970 spa_t *spa; 1971 nvlist_t *config, *props = NULL; 1972 int error; 1973 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT); 1974 1975 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 1976 return (error); 1977 1978 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size, 1979 zc->zc_iflags, &config)) { 1980 spa_close(spa, FTAG); 1981 return (error); 1982 } 1983 1984 if (zc->zc_nvlist_src_size != 0 && (error = 1985 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 1986 zc->zc_iflags, &props))) { 1987 spa_close(spa, FTAG); 1988 nvlist_free(config); 1989 return (error); 1990 } 1991 1992 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp); 1993 1994 spa_close(spa, FTAG); 1995 1996 nvlist_free(config); 1997 nvlist_free(props); 1998 1999 return (error); 2000 } 2001 2002 static int 2003 zfs_ioc_vdev_setpath(zfs_cmd_t *zc) 2004 { 2005 spa_t *spa; 2006 char *path = zc->zc_value; 2007 uint64_t guid = zc->zc_guid; 2008 int error; 2009 2010 error = spa_open(zc->zc_name, &spa, FTAG); 2011 if (error != 0) 2012 return (error); 2013 2014 error = spa_vdev_setpath(spa, guid, path); 2015 spa_close(spa, FTAG); 2016 return (error); 2017 } 2018 2019 static int 2020 zfs_ioc_vdev_setfru(zfs_cmd_t *zc) 2021 { 2022 spa_t *spa; 2023 char *fru = zc->zc_value; 2024 uint64_t guid = zc->zc_guid; 2025 int error; 2026 2027 error = spa_open(zc->zc_name, &spa, FTAG); 2028 if (error != 0) 2029 return (error); 2030 2031 error = spa_vdev_setfru(spa, guid, fru); 2032 spa_close(spa, FTAG); 2033 return (error); 2034 } 2035 2036 static int 2037 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os) 2038 { 2039 int error = 0; 2040 nvlist_t *nv; 2041 2042 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2043 2044 if (zc->zc_nvlist_dst != 0 && 2045 (error = dsl_prop_get_all(os, &nv)) == 0) { 2046 dmu_objset_stats(os, nv); 2047 /* 2048 * NB: zvol_get_stats() will read the objset contents, 2049 * which we aren't supposed to do with a 2050 * DS_MODE_USER hold, because it could be 2051 * inconsistent. So this is a bit of a workaround... 2052 * XXX reading with out owning 2053 */ 2054 if (!zc->zc_objset_stats.dds_inconsistent && 2055 dmu_objset_type(os) == DMU_OST_ZVOL) { 2056 error = zvol_get_stats(os, nv); 2057 if (error == EIO) 2058 return (error); 2059 VERIFY0(error); 2060 } 2061 error = put_nvlist(zc, nv); 2062 nvlist_free(nv); 2063 } 2064 2065 return (error); 2066 } 2067 2068 /* 2069 * inputs: 2070 * zc_name name of filesystem 2071 * zc_nvlist_dst_size size of buffer for property nvlist 2072 * 2073 * outputs: 2074 * zc_objset_stats stats 2075 * zc_nvlist_dst property nvlist 2076 * zc_nvlist_dst_size size of property nvlist 2077 */ 2078 static int 2079 zfs_ioc_objset_stats(zfs_cmd_t *zc) 2080 { 2081 objset_t *os; 2082 int error; 2083 2084 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2085 if (error == 0) { 2086 error = zfs_ioc_objset_stats_impl(zc, os); 2087 dmu_objset_rele(os, FTAG); 2088 } 2089 2090 return (error); 2091 } 2092 2093 /* 2094 * inputs: 2095 * zc_name name of filesystem 2096 * zc_nvlist_dst_size size of buffer for property nvlist 2097 * 2098 * outputs: 2099 * zc_nvlist_dst received property nvlist 2100 * zc_nvlist_dst_size size of received property nvlist 2101 * 2102 * Gets received properties (distinct from local properties on or after 2103 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from 2104 * local property values. 2105 */ 2106 static int 2107 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc) 2108 { 2109 int error = 0; 2110 nvlist_t *nv; 2111 2112 /* 2113 * Without this check, we would return local property values if the 2114 * caller has not already received properties on or after 2115 * SPA_VERSION_RECVD_PROPS. 2116 */ 2117 if (!dsl_prop_get_hasrecvd(zc->zc_name)) 2118 return (SET_ERROR(ENOTSUP)); 2119 2120 if (zc->zc_nvlist_dst != 0 && 2121 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) { 2122 error = put_nvlist(zc, nv); 2123 nvlist_free(nv); 2124 } 2125 2126 return (error); 2127 } 2128 2129 static int 2130 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) 2131 { 2132 uint64_t value; 2133 int error; 2134 2135 /* 2136 * zfs_get_zplprop() will either find a value or give us 2137 * the default value (if there is one). 2138 */ 2139 if ((error = zfs_get_zplprop(os, prop, &value)) != 0) 2140 return (error); 2141 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); 2142 return (0); 2143 } 2144 2145 /* 2146 * inputs: 2147 * zc_name name of filesystem 2148 * zc_nvlist_dst_size size of buffer for zpl property nvlist 2149 * 2150 * outputs: 2151 * zc_nvlist_dst zpl property nvlist 2152 * zc_nvlist_dst_size size of zpl property nvlist 2153 */ 2154 static int 2155 zfs_ioc_objset_zplprops(zfs_cmd_t *zc) 2156 { 2157 objset_t *os; 2158 int err; 2159 2160 /* XXX reading without owning */ 2161 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os)) 2162 return (err); 2163 2164 dmu_objset_fast_stat(os, &zc->zc_objset_stats); 2165 2166 /* 2167 * NB: nvl_add_zplprop() will read the objset contents, 2168 * which we aren't supposed to do with a DS_MODE_USER 2169 * hold, because it could be inconsistent. 2170 */ 2171 if (zc->zc_nvlist_dst != NULL && 2172 !zc->zc_objset_stats.dds_inconsistent && 2173 dmu_objset_type(os) == DMU_OST_ZFS) { 2174 nvlist_t *nv; 2175 2176 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2177 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && 2178 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && 2179 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && 2180 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0) 2181 err = put_nvlist(zc, nv); 2182 nvlist_free(nv); 2183 } else { 2184 err = SET_ERROR(ENOENT); 2185 } 2186 dmu_objset_rele(os, FTAG); 2187 return (err); 2188 } 2189 2190 static boolean_t 2191 dataset_name_hidden(const char *name) 2192 { 2193 /* 2194 * Skip over datasets that are not visible in this zone, 2195 * internal datasets (which have a $ in their name), and 2196 * temporary datasets (which have a % in their name). 2197 */ 2198 if (strchr(name, '$') != NULL) 2199 return (B_TRUE); 2200 if (strchr(name, '%') != NULL) 2201 return (B_TRUE); 2202 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL)) 2203 return (B_TRUE); 2204 return (B_FALSE); 2205 } 2206 2207 /* 2208 * inputs: 2209 * zc_name name of filesystem 2210 * zc_cookie zap cursor 2211 * zc_nvlist_dst_size size of buffer for property nvlist 2212 * 2213 * outputs: 2214 * zc_name name of next filesystem 2215 * zc_cookie zap cursor 2216 * zc_objset_stats stats 2217 * zc_nvlist_dst property nvlist 2218 * zc_nvlist_dst_size size of property nvlist 2219 */ 2220 static int 2221 zfs_ioc_dataset_list_next(zfs_cmd_t *zc) 2222 { 2223 objset_t *os; 2224 int error; 2225 char *p; 2226 size_t orig_len = strlen(zc->zc_name); 2227 2228 top: 2229 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) { 2230 if (error == ENOENT) 2231 error = SET_ERROR(ESRCH); 2232 return (error); 2233 } 2234 2235 p = strrchr(zc->zc_name, '/'); 2236 if (p == NULL || p[1] != '\0') 2237 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name)); 2238 p = zc->zc_name + strlen(zc->zc_name); 2239 2240 do { 2241 error = dmu_dir_list_next(os, 2242 sizeof (zc->zc_name) - (p - zc->zc_name), p, 2243 NULL, &zc->zc_cookie); 2244 if (error == ENOENT) 2245 error = SET_ERROR(ESRCH); 2246 } while (error == 0 && dataset_name_hidden(zc->zc_name)); 2247 dmu_objset_rele(os, FTAG); 2248 2249 /* 2250 * If it's an internal dataset (ie. with a '$' in its name), 2251 * don't try to get stats for it, otherwise we'll return ENOENT. 2252 */ 2253 if (error == 0 && strchr(zc->zc_name, '$') == NULL) { 2254 error = zfs_ioc_objset_stats(zc); /* fill in the stats */ 2255 if (error == ENOENT) { 2256 /* We lost a race with destroy, get the next one. */ 2257 zc->zc_name[orig_len] = '\0'; 2258 goto top; 2259 } 2260 } 2261 return (error); 2262 } 2263 2264 /* 2265 * inputs: 2266 * zc_name name of filesystem 2267 * zc_cookie zap cursor 2268 * zc_nvlist_dst_size size of buffer for property nvlist 2269 * 2270 * outputs: 2271 * zc_name name of next snapshot 2272 * zc_objset_stats stats 2273 * zc_nvlist_dst property nvlist 2274 * zc_nvlist_dst_size size of property nvlist 2275 */ 2276 static int 2277 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc) 2278 { 2279 objset_t *os; 2280 int error; 2281 2282 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 2283 if (error != 0) { 2284 return (error == ENOENT ? ESRCH : error); 2285 } 2286 2287 /* 2288 * A dataset name of maximum length cannot have any snapshots, 2289 * so exit immediately. 2290 */ 2291 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) { 2292 dmu_objset_rele(os, FTAG); 2293 return (SET_ERROR(ESRCH)); 2294 } 2295 2296 error = dmu_snapshot_list_next(os, 2297 sizeof (zc->zc_name) - strlen(zc->zc_name), 2298 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie, 2299 NULL); 2300 2301 if (error == 0) { 2302 dsl_dataset_t *ds; 2303 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 2304 2305 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds); 2306 if (error == 0) { 2307 objset_t *ossnap; 2308 2309 error = dmu_objset_from_ds(ds, &ossnap); 2310 if (error == 0) 2311 error = zfs_ioc_objset_stats_impl(zc, ossnap); 2312 dsl_dataset_rele(ds, FTAG); 2313 } 2314 } else if (error == ENOENT) { 2315 error = SET_ERROR(ESRCH); 2316 } 2317 2318 dmu_objset_rele(os, FTAG); 2319 /* if we failed, undo the @ that we tacked on to zc_name */ 2320 if (error != 0) 2321 *strchr(zc->zc_name, '@') = '\0'; 2322 return (error); 2323 } 2324 2325 static int 2326 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) 2327 { 2328 const char *propname = nvpair_name(pair); 2329 uint64_t *valary; 2330 unsigned int vallen; 2331 const char *domain; 2332 char *dash; 2333 zfs_userquota_prop_t type; 2334 uint64_t rid; 2335 uint64_t quota; 2336 zfsvfs_t *zfsvfs; 2337 int err; 2338 2339 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2340 nvlist_t *attrs; 2341 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2342 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2343 &pair) != 0) 2344 return (SET_ERROR(EINVAL)); 2345 } 2346 2347 /* 2348 * A correctly constructed propname is encoded as 2349 * userquota@<rid>-<domain>. 2350 */ 2351 if ((dash = strchr(propname, '-')) == NULL || 2352 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 || 2353 vallen != 3) 2354 return (SET_ERROR(EINVAL)); 2355 2356 domain = dash + 1; 2357 type = valary[0]; 2358 rid = valary[1]; 2359 quota = valary[2]; 2360 2361 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE); 2362 if (err == 0) { 2363 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota); 2364 zfsvfs_rele(zfsvfs, FTAG); 2365 } 2366 2367 return (err); 2368 } 2369 2370 /* 2371 * If the named property is one that has a special function to set its value, 2372 * return 0 on success and a positive error code on failure; otherwise if it is 2373 * not one of the special properties handled by this function, return -1. 2374 * 2375 * XXX: It would be better for callers of the property interface if we handled 2376 * these special cases in dsl_prop.c (in the dsl layer). 2377 */ 2378 static int 2379 zfs_prop_set_special(const char *dsname, zprop_source_t source, 2380 nvpair_t *pair) 2381 { 2382 const char *propname = nvpair_name(pair); 2383 zfs_prop_t prop = zfs_name_to_prop(propname); 2384 uint64_t intval; 2385 int err = -1; 2386 2387 if (prop == ZPROP_INVAL) { 2388 if (zfs_prop_userquota(propname)) 2389 return (zfs_prop_set_userquota(dsname, pair)); 2390 return (-1); 2391 } 2392 2393 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2394 nvlist_t *attrs; 2395 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 2396 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2397 &pair) == 0); 2398 } 2399 2400 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) 2401 return (-1); 2402 2403 VERIFY(0 == nvpair_value_uint64(pair, &intval)); 2404 2405 switch (prop) { 2406 case ZFS_PROP_QUOTA: 2407 err = dsl_dir_set_quota(dsname, source, intval); 2408 break; 2409 case ZFS_PROP_REFQUOTA: 2410 err = dsl_dataset_set_refquota(dsname, source, intval); 2411 break; 2412 case ZFS_PROP_FILESYSTEM_LIMIT: 2413 case ZFS_PROP_SNAPSHOT_LIMIT: 2414 if (intval == UINT64_MAX) { 2415 /* clearing the limit, just do it */ 2416 err = 0; 2417 } else { 2418 err = dsl_dir_activate_fs_ss_limit(dsname); 2419 } 2420 /* 2421 * Set err to -1 to force the zfs_set_prop_nvlist code down the 2422 * default path to set the value in the nvlist. 2423 */ 2424 if (err == 0) 2425 err = -1; 2426 break; 2427 case ZFS_PROP_RESERVATION: 2428 err = dsl_dir_set_reservation(dsname, source, intval); 2429 break; 2430 case ZFS_PROP_REFRESERVATION: 2431 err = dsl_dataset_set_refreservation(dsname, source, intval); 2432 break; 2433 case ZFS_PROP_VOLSIZE: 2434 err = zvol_set_volsize(dsname, intval); 2435 break; 2436 case ZFS_PROP_VERSION: 2437 { 2438 zfsvfs_t *zfsvfs; 2439 2440 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0) 2441 break; 2442 2443 err = zfs_set_version(zfsvfs, intval); 2444 zfsvfs_rele(zfsvfs, FTAG); 2445 2446 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) { 2447 zfs_cmd_t *zc; 2448 2449 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 2450 (void) strcpy(zc->zc_name, dsname); 2451 (void) zfs_ioc_userspace_upgrade(zc); 2452 kmem_free(zc, sizeof (zfs_cmd_t)); 2453 } 2454 break; 2455 } 2456 default: 2457 err = -1; 2458 } 2459 2460 return (err); 2461 } 2462 2463 /* 2464 * This function is best effort. If it fails to set any of the given properties, 2465 * it continues to set as many as it can and returns the last error 2466 * encountered. If the caller provides a non-NULL errlist, it will be filled in 2467 * with the list of names of all the properties that failed along with the 2468 * corresponding error numbers. 2469 * 2470 * If every property is set successfully, zero is returned and errlist is not 2471 * modified. 2472 */ 2473 int 2474 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl, 2475 nvlist_t *errlist) 2476 { 2477 nvpair_t *pair; 2478 nvpair_t *propval; 2479 int rv = 0; 2480 uint64_t intval; 2481 char *strval; 2482 nvlist_t *genericnvl = fnvlist_alloc(); 2483 nvlist_t *retrynvl = fnvlist_alloc(); 2484 2485 retry: 2486 pair = NULL; 2487 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2488 const char *propname = nvpair_name(pair); 2489 zfs_prop_t prop = zfs_name_to_prop(propname); 2490 int err = 0; 2491 2492 /* decode the property value */ 2493 propval = pair; 2494 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2495 nvlist_t *attrs; 2496 attrs = fnvpair_value_nvlist(pair); 2497 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 2498 &propval) != 0) 2499 err = SET_ERROR(EINVAL); 2500 } 2501 2502 /* Validate value type */ 2503 if (err == 0 && prop == ZPROP_INVAL) { 2504 if (zfs_prop_user(propname)) { 2505 if (nvpair_type(propval) != DATA_TYPE_STRING) 2506 err = SET_ERROR(EINVAL); 2507 } else if (zfs_prop_userquota(propname)) { 2508 if (nvpair_type(propval) != 2509 DATA_TYPE_UINT64_ARRAY) 2510 err = SET_ERROR(EINVAL); 2511 } else { 2512 err = SET_ERROR(EINVAL); 2513 } 2514 } else if (err == 0) { 2515 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2516 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING) 2517 err = SET_ERROR(EINVAL); 2518 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) { 2519 const char *unused; 2520 2521 intval = fnvpair_value_uint64(propval); 2522 2523 switch (zfs_prop_get_type(prop)) { 2524 case PROP_TYPE_NUMBER: 2525 break; 2526 case PROP_TYPE_STRING: 2527 err = SET_ERROR(EINVAL); 2528 break; 2529 case PROP_TYPE_INDEX: 2530 if (zfs_prop_index_to_string(prop, 2531 intval, &unused) != 0) 2532 err = SET_ERROR(EINVAL); 2533 break; 2534 default: 2535 cmn_err(CE_PANIC, 2536 "unknown property type"); 2537 } 2538 } else { 2539 err = SET_ERROR(EINVAL); 2540 } 2541 } 2542 2543 /* Validate permissions */ 2544 if (err == 0) 2545 err = zfs_check_settable(dsname, pair, CRED()); 2546 2547 if (err == 0) { 2548 err = zfs_prop_set_special(dsname, source, pair); 2549 if (err == -1) { 2550 /* 2551 * For better performance we build up a list of 2552 * properties to set in a single transaction. 2553 */ 2554 err = nvlist_add_nvpair(genericnvl, pair); 2555 } else if (err != 0 && nvl != retrynvl) { 2556 /* 2557 * This may be a spurious error caused by 2558 * receiving quota and reservation out of order. 2559 * Try again in a second pass. 2560 */ 2561 err = nvlist_add_nvpair(retrynvl, pair); 2562 } 2563 } 2564 2565 if (err != 0) { 2566 if (errlist != NULL) 2567 fnvlist_add_int32(errlist, propname, err); 2568 rv = err; 2569 } 2570 } 2571 2572 if (nvl != retrynvl && !nvlist_empty(retrynvl)) { 2573 nvl = retrynvl; 2574 goto retry; 2575 } 2576 2577 if (!nvlist_empty(genericnvl) && 2578 dsl_props_set(dsname, source, genericnvl) != 0) { 2579 /* 2580 * If this fails, we still want to set as many properties as we 2581 * can, so try setting them individually. 2582 */ 2583 pair = NULL; 2584 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) { 2585 const char *propname = nvpair_name(pair); 2586 int err = 0; 2587 2588 propval = pair; 2589 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 2590 nvlist_t *attrs; 2591 attrs = fnvpair_value_nvlist(pair); 2592 propval = fnvlist_lookup_nvpair(attrs, 2593 ZPROP_VALUE); 2594 } 2595 2596 if (nvpair_type(propval) == DATA_TYPE_STRING) { 2597 strval = fnvpair_value_string(propval); 2598 err = dsl_prop_set_string(dsname, propname, 2599 source, strval); 2600 } else { 2601 intval = fnvpair_value_uint64(propval); 2602 err = dsl_prop_set_int(dsname, propname, source, 2603 intval); 2604 } 2605 2606 if (err != 0) { 2607 if (errlist != NULL) { 2608 fnvlist_add_int32(errlist, propname, 2609 err); 2610 } 2611 rv = err; 2612 } 2613 } 2614 } 2615 nvlist_free(genericnvl); 2616 nvlist_free(retrynvl); 2617 2618 return (rv); 2619 } 2620 2621 /* 2622 * Check that all the properties are valid user properties. 2623 */ 2624 static int 2625 zfs_check_userprops(const char *fsname, nvlist_t *nvl) 2626 { 2627 nvpair_t *pair = NULL; 2628 int error = 0; 2629 2630 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) { 2631 const char *propname = nvpair_name(pair); 2632 2633 if (!zfs_prop_user(propname) || 2634 nvpair_type(pair) != DATA_TYPE_STRING) 2635 return (SET_ERROR(EINVAL)); 2636 2637 if (error = zfs_secpolicy_write_perms(fsname, 2638 ZFS_DELEG_PERM_USERPROP, CRED())) 2639 return (error); 2640 2641 if (strlen(propname) >= ZAP_MAXNAMELEN) 2642 return (SET_ERROR(ENAMETOOLONG)); 2643 2644 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN) 2645 return (E2BIG); 2646 } 2647 return (0); 2648 } 2649 2650 static void 2651 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) 2652 { 2653 nvpair_t *pair; 2654 2655 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2656 2657 pair = NULL; 2658 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { 2659 if (nvlist_exists(skipped, nvpair_name(pair))) 2660 continue; 2661 2662 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); 2663 } 2664 } 2665 2666 static int 2667 clear_received_props(const char *dsname, nvlist_t *props, 2668 nvlist_t *skipped) 2669 { 2670 int err = 0; 2671 nvlist_t *cleared_props = NULL; 2672 props_skip(props, skipped, &cleared_props); 2673 if (!nvlist_empty(cleared_props)) { 2674 /* 2675 * Acts on local properties until the dataset has received 2676 * properties at least once on or after SPA_VERSION_RECVD_PROPS. 2677 */ 2678 zprop_source_t flags = (ZPROP_SRC_NONE | 2679 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0)); 2680 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL); 2681 } 2682 nvlist_free(cleared_props); 2683 return (err); 2684 } 2685 2686 /* 2687 * inputs: 2688 * zc_name name of filesystem 2689 * zc_value name of property to set 2690 * zc_nvlist_src{_size} nvlist of properties to apply 2691 * zc_cookie received properties flag 2692 * 2693 * outputs: 2694 * zc_nvlist_dst{_size} error for each unapplied received property 2695 */ 2696 static int 2697 zfs_ioc_set_prop(zfs_cmd_t *zc) 2698 { 2699 nvlist_t *nvl; 2700 boolean_t received = zc->zc_cookie; 2701 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED : 2702 ZPROP_SRC_LOCAL); 2703 nvlist_t *errors; 2704 int error; 2705 2706 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2707 zc->zc_iflags, &nvl)) != 0) 2708 return (error); 2709 2710 if (received) { 2711 nvlist_t *origprops; 2712 2713 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) { 2714 (void) clear_received_props(zc->zc_name, 2715 origprops, nvl); 2716 nvlist_free(origprops); 2717 } 2718 2719 error = dsl_prop_set_hasrecvd(zc->zc_name); 2720 } 2721 2722 errors = fnvlist_alloc(); 2723 if (error == 0) 2724 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors); 2725 2726 if (zc->zc_nvlist_dst != NULL && errors != NULL) { 2727 (void) put_nvlist(zc, errors); 2728 } 2729 2730 nvlist_free(errors); 2731 nvlist_free(nvl); 2732 return (error); 2733 } 2734 2735 /* 2736 * inputs: 2737 * zc_name name of filesystem 2738 * zc_value name of property to inherit 2739 * zc_cookie revert to received value if TRUE 2740 * 2741 * outputs: none 2742 */ 2743 static int 2744 zfs_ioc_inherit_prop(zfs_cmd_t *zc) 2745 { 2746 const char *propname = zc->zc_value; 2747 zfs_prop_t prop = zfs_name_to_prop(propname); 2748 boolean_t received = zc->zc_cookie; 2749 zprop_source_t source = (received 2750 ? ZPROP_SRC_NONE /* revert to received value, if any */ 2751 : ZPROP_SRC_INHERITED); /* explicitly inherit */ 2752 2753 if (received) { 2754 nvlist_t *dummy; 2755 nvpair_t *pair; 2756 zprop_type_t type; 2757 int err; 2758 2759 /* 2760 * zfs_prop_set_special() expects properties in the form of an 2761 * nvpair with type info. 2762 */ 2763 if (prop == ZPROP_INVAL) { 2764 if (!zfs_prop_user(propname)) 2765 return (SET_ERROR(EINVAL)); 2766 2767 type = PROP_TYPE_STRING; 2768 } else if (prop == ZFS_PROP_VOLSIZE || 2769 prop == ZFS_PROP_VERSION) { 2770 return (SET_ERROR(EINVAL)); 2771 } else { 2772 type = zfs_prop_get_type(prop); 2773 } 2774 2775 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2776 2777 switch (type) { 2778 case PROP_TYPE_STRING: 2779 VERIFY(0 == nvlist_add_string(dummy, propname, "")); 2780 break; 2781 case PROP_TYPE_NUMBER: 2782 case PROP_TYPE_INDEX: 2783 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); 2784 break; 2785 default: 2786 nvlist_free(dummy); 2787 return (SET_ERROR(EINVAL)); 2788 } 2789 2790 pair = nvlist_next_nvpair(dummy, NULL); 2791 err = zfs_prop_set_special(zc->zc_name, source, pair); 2792 nvlist_free(dummy); 2793 if (err != -1) 2794 return (err); /* special property already handled */ 2795 } else { 2796 /* 2797 * Only check this in the non-received case. We want to allow 2798 * 'inherit -S' to revert non-inheritable properties like quota 2799 * and reservation to the received or default values even though 2800 * they are not considered inheritable. 2801 */ 2802 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop)) 2803 return (SET_ERROR(EINVAL)); 2804 } 2805 2806 /* property name has been validated by zfs_secpolicy_inherit_prop() */ 2807 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source)); 2808 } 2809 2810 static int 2811 zfs_ioc_pool_set_props(zfs_cmd_t *zc) 2812 { 2813 nvlist_t *props; 2814 spa_t *spa; 2815 int error; 2816 nvpair_t *pair; 2817 2818 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2819 zc->zc_iflags, &props)) 2820 return (error); 2821 2822 /* 2823 * If the only property is the configfile, then just do a spa_lookup() 2824 * to handle the faulted case. 2825 */ 2826 pair = nvlist_next_nvpair(props, NULL); 2827 if (pair != NULL && strcmp(nvpair_name(pair), 2828 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 && 2829 nvlist_next_nvpair(props, pair) == NULL) { 2830 mutex_enter(&spa_namespace_lock); 2831 if ((spa = spa_lookup(zc->zc_name)) != NULL) { 2832 spa_configfile_set(spa, props, B_FALSE); 2833 spa_config_sync(spa, B_FALSE, B_TRUE); 2834 } 2835 mutex_exit(&spa_namespace_lock); 2836 if (spa != NULL) { 2837 nvlist_free(props); 2838 return (0); 2839 } 2840 } 2841 2842 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2843 nvlist_free(props); 2844 return (error); 2845 } 2846 2847 error = spa_prop_set(spa, props); 2848 2849 nvlist_free(props); 2850 spa_close(spa, FTAG); 2851 2852 return (error); 2853 } 2854 2855 static int 2856 zfs_ioc_pool_get_props(zfs_cmd_t *zc) 2857 { 2858 spa_t *spa; 2859 int error; 2860 nvlist_t *nvp = NULL; 2861 2862 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) { 2863 /* 2864 * If the pool is faulted, there may be properties we can still 2865 * get (such as altroot and cachefile), so attempt to get them 2866 * anyway. 2867 */ 2868 mutex_enter(&spa_namespace_lock); 2869 if ((spa = spa_lookup(zc->zc_name)) != NULL) 2870 error = spa_prop_get(spa, &nvp); 2871 mutex_exit(&spa_namespace_lock); 2872 } else { 2873 error = spa_prop_get(spa, &nvp); 2874 spa_close(spa, FTAG); 2875 } 2876 2877 if (error == 0 && zc->zc_nvlist_dst != NULL) 2878 error = put_nvlist(zc, nvp); 2879 else 2880 error = SET_ERROR(EFAULT); 2881 2882 nvlist_free(nvp); 2883 return (error); 2884 } 2885 2886 /* 2887 * inputs: 2888 * zc_name name of filesystem 2889 * zc_nvlist_src{_size} nvlist of delegated permissions 2890 * zc_perm_action allow/unallow flag 2891 * 2892 * outputs: none 2893 */ 2894 static int 2895 zfs_ioc_set_fsacl(zfs_cmd_t *zc) 2896 { 2897 int error; 2898 nvlist_t *fsaclnv = NULL; 2899 2900 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 2901 zc->zc_iflags, &fsaclnv)) != 0) 2902 return (error); 2903 2904 /* 2905 * Verify nvlist is constructed correctly 2906 */ 2907 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) { 2908 nvlist_free(fsaclnv); 2909 return (SET_ERROR(EINVAL)); 2910 } 2911 2912 /* 2913 * If we don't have PRIV_SYS_MOUNT, then validate 2914 * that user is allowed to hand out each permission in 2915 * the nvlist(s) 2916 */ 2917 2918 error = secpolicy_zfs(CRED()); 2919 if (error != 0) { 2920 if (zc->zc_perm_action == B_FALSE) { 2921 error = dsl_deleg_can_allow(zc->zc_name, 2922 fsaclnv, CRED()); 2923 } else { 2924 error = dsl_deleg_can_unallow(zc->zc_name, 2925 fsaclnv, CRED()); 2926 } 2927 } 2928 2929 if (error == 0) 2930 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action); 2931 2932 nvlist_free(fsaclnv); 2933 return (error); 2934 } 2935 2936 /* 2937 * inputs: 2938 * zc_name name of filesystem 2939 * 2940 * outputs: 2941 * zc_nvlist_src{_size} nvlist of delegated permissions 2942 */ 2943 static int 2944 zfs_ioc_get_fsacl(zfs_cmd_t *zc) 2945 { 2946 nvlist_t *nvp; 2947 int error; 2948 2949 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) { 2950 error = put_nvlist(zc, nvp); 2951 nvlist_free(nvp); 2952 } 2953 2954 return (error); 2955 } 2956 2957 /* 2958 * Search the vfs list for a specified resource. Returns a pointer to it 2959 * or NULL if no suitable entry is found. The caller of this routine 2960 * is responsible for releasing the returned vfs pointer. 2961 */ 2962 static vfs_t * 2963 zfs_get_vfs(const char *resource) 2964 { 2965 struct vfs *vfsp; 2966 struct vfs *vfs_found = NULL; 2967 2968 vfs_list_read_lock(); 2969 vfsp = rootvfs; 2970 do { 2971 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) { 2972 VFS_HOLD(vfsp); 2973 vfs_found = vfsp; 2974 break; 2975 } 2976 vfsp = vfsp->vfs_next; 2977 } while (vfsp != rootvfs); 2978 vfs_list_unlock(); 2979 return (vfs_found); 2980 } 2981 2982 /* ARGSUSED */ 2983 static void 2984 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 2985 { 2986 zfs_creat_t *zct = arg; 2987 2988 zfs_create_fs(os, cr, zct->zct_zplprops, tx); 2989 } 2990 2991 #define ZFS_PROP_UNDEFINED ((uint64_t)-1) 2992 2993 /* 2994 * inputs: 2995 * os parent objset pointer (NULL if root fs) 2996 * fuids_ok fuids allowed in this version of the spa? 2997 * sa_ok SAs allowed in this version of the spa? 2998 * createprops list of properties requested by creator 2999 * 3000 * outputs: 3001 * zplprops values for the zplprops we attach to the master node object 3002 * is_ci true if requested file system will be purely case-insensitive 3003 * 3004 * Determine the settings for utf8only, normalization and 3005 * casesensitivity. Specific values may have been requested by the 3006 * creator and/or we can inherit values from the parent dataset. If 3007 * the file system is of too early a vintage, a creator can not 3008 * request settings for these properties, even if the requested 3009 * setting is the default value. We don't actually want to create dsl 3010 * properties for these, so remove them from the source nvlist after 3011 * processing. 3012 */ 3013 static int 3014 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, 3015 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops, 3016 nvlist_t *zplprops, boolean_t *is_ci) 3017 { 3018 uint64_t sense = ZFS_PROP_UNDEFINED; 3019 uint64_t norm = ZFS_PROP_UNDEFINED; 3020 uint64_t u8 = ZFS_PROP_UNDEFINED; 3021 3022 ASSERT(zplprops != NULL); 3023 3024 /* 3025 * Pull out creator prop choices, if any. 3026 */ 3027 if (createprops) { 3028 (void) nvlist_lookup_uint64(createprops, 3029 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver); 3030 (void) nvlist_lookup_uint64(createprops, 3031 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm); 3032 (void) nvlist_remove_all(createprops, 3033 zfs_prop_to_name(ZFS_PROP_NORMALIZE)); 3034 (void) nvlist_lookup_uint64(createprops, 3035 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8); 3036 (void) nvlist_remove_all(createprops, 3037 zfs_prop_to_name(ZFS_PROP_UTF8ONLY)); 3038 (void) nvlist_lookup_uint64(createprops, 3039 zfs_prop_to_name(ZFS_PROP_CASE), &sense); 3040 (void) nvlist_remove_all(createprops, 3041 zfs_prop_to_name(ZFS_PROP_CASE)); 3042 } 3043 3044 /* 3045 * If the zpl version requested is whacky or the file system 3046 * or pool is version is too "young" to support normalization 3047 * and the creator tried to set a value for one of the props, 3048 * error out. 3049 */ 3050 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) || 3051 (zplver >= ZPL_VERSION_FUID && !fuids_ok) || 3052 (zplver >= ZPL_VERSION_SA && !sa_ok) || 3053 (zplver < ZPL_VERSION_NORMALIZATION && 3054 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED || 3055 sense != ZFS_PROP_UNDEFINED))) 3056 return (SET_ERROR(ENOTSUP)); 3057 3058 /* 3059 * Put the version in the zplprops 3060 */ 3061 VERIFY(nvlist_add_uint64(zplprops, 3062 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); 3063 3064 if (norm == ZFS_PROP_UNDEFINED) 3065 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0); 3066 VERIFY(nvlist_add_uint64(zplprops, 3067 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); 3068 3069 /* 3070 * If we're normalizing, names must always be valid UTF-8 strings. 3071 */ 3072 if (norm) 3073 u8 = 1; 3074 if (u8 == ZFS_PROP_UNDEFINED) 3075 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0); 3076 VERIFY(nvlist_add_uint64(zplprops, 3077 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); 3078 3079 if (sense == ZFS_PROP_UNDEFINED) 3080 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0); 3081 VERIFY(nvlist_add_uint64(zplprops, 3082 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); 3083 3084 if (is_ci) 3085 *is_ci = (sense == ZFS_CASE_INSENSITIVE); 3086 3087 return (0); 3088 } 3089 3090 static int 3091 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops, 3092 nvlist_t *zplprops, boolean_t *is_ci) 3093 { 3094 boolean_t fuids_ok, sa_ok; 3095 uint64_t zplver = ZPL_VERSION; 3096 objset_t *os = NULL; 3097 char parentname[MAXNAMELEN]; 3098 char *cp; 3099 spa_t *spa; 3100 uint64_t spa_vers; 3101 int error; 3102 3103 (void) strlcpy(parentname, dataset, sizeof (parentname)); 3104 cp = strrchr(parentname, '/'); 3105 ASSERT(cp != NULL); 3106 cp[0] = '\0'; 3107 3108 if ((error = spa_open(dataset, &spa, FTAG)) != 0) 3109 return (error); 3110 3111 spa_vers = spa_version(spa); 3112 spa_close(spa, FTAG); 3113 3114 zplver = zfs_zpl_version_map(spa_vers); 3115 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3116 sa_ok = (zplver >= ZPL_VERSION_SA); 3117 3118 /* 3119 * Open parent object set so we can inherit zplprop values. 3120 */ 3121 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0) 3122 return (error); 3123 3124 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops, 3125 zplprops, is_ci); 3126 dmu_objset_rele(os, FTAG); 3127 return (error); 3128 } 3129 3130 static int 3131 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops, 3132 nvlist_t *zplprops, boolean_t *is_ci) 3133 { 3134 boolean_t fuids_ok; 3135 boolean_t sa_ok; 3136 uint64_t zplver = ZPL_VERSION; 3137 int error; 3138 3139 zplver = zfs_zpl_version_map(spa_vers); 3140 fuids_ok = (zplver >= ZPL_VERSION_FUID); 3141 sa_ok = (zplver >= ZPL_VERSION_SA); 3142 3143 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok, 3144 createprops, zplprops, is_ci); 3145 return (error); 3146 } 3147 3148 /* 3149 * innvl: { 3150 * "type" -> dmu_objset_type_t (int32) 3151 * (optional) "props" -> { prop -> value } 3152 * } 3153 * 3154 * outnvl: propname -> error code (int32) 3155 */ 3156 static int 3157 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3158 { 3159 int error = 0; 3160 zfs_creat_t zct = { 0 }; 3161 nvlist_t *nvprops = NULL; 3162 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); 3163 int32_t type32; 3164 dmu_objset_type_t type; 3165 boolean_t is_insensitive = B_FALSE; 3166 3167 if (nvlist_lookup_int32(innvl, "type", &type32) != 0) 3168 return (SET_ERROR(EINVAL)); 3169 type = type32; 3170 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3171 3172 switch (type) { 3173 case DMU_OST_ZFS: 3174 cbfunc = zfs_create_cb; 3175 break; 3176 3177 case DMU_OST_ZVOL: 3178 cbfunc = zvol_create_cb; 3179 break; 3180 3181 default: 3182 cbfunc = NULL; 3183 break; 3184 } 3185 if (strchr(fsname, '@') || 3186 strchr(fsname, '%')) 3187 return (SET_ERROR(EINVAL)); 3188 3189 zct.zct_props = nvprops; 3190 3191 if (cbfunc == NULL) 3192 return (SET_ERROR(EINVAL)); 3193 3194 if (type == DMU_OST_ZVOL) { 3195 uint64_t volsize, volblocksize; 3196 3197 if (nvprops == NULL) 3198 return (SET_ERROR(EINVAL)); 3199 if (nvlist_lookup_uint64(nvprops, 3200 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0) 3201 return (SET_ERROR(EINVAL)); 3202 3203 if ((error = nvlist_lookup_uint64(nvprops, 3204 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 3205 &volblocksize)) != 0 && error != ENOENT) 3206 return (SET_ERROR(EINVAL)); 3207 3208 if (error != 0) 3209 volblocksize = zfs_prop_default_numeric( 3210 ZFS_PROP_VOLBLOCKSIZE); 3211 3212 if ((error = zvol_check_volblocksize( 3213 volblocksize)) != 0 || 3214 (error = zvol_check_volsize(volsize, 3215 volblocksize)) != 0) 3216 return (error); 3217 } else if (type == DMU_OST_ZFS) { 3218 int error; 3219 3220 /* 3221 * We have to have normalization and 3222 * case-folding flags correct when we do the 3223 * file system creation, so go figure them out 3224 * now. 3225 */ 3226 VERIFY(nvlist_alloc(&zct.zct_zplprops, 3227 NV_UNIQUE_NAME, KM_SLEEP) == 0); 3228 error = zfs_fill_zplprops(fsname, nvprops, 3229 zct.zct_zplprops, &is_insensitive); 3230 if (error != 0) { 3231 nvlist_free(zct.zct_zplprops); 3232 return (error); 3233 } 3234 } 3235 3236 error = dmu_objset_create(fsname, type, 3237 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct); 3238 nvlist_free(zct.zct_zplprops); 3239 3240 /* 3241 * It would be nice to do this atomically. 3242 */ 3243 if (error == 0) { 3244 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3245 nvprops, outnvl); 3246 if (error != 0) 3247 (void) dsl_destroy_head(fsname); 3248 } 3249 return (error); 3250 } 3251 3252 /* 3253 * innvl: { 3254 * "origin" -> name of origin snapshot 3255 * (optional) "props" -> { prop -> value } 3256 * } 3257 * 3258 * outnvl: propname -> error code (int32) 3259 */ 3260 static int 3261 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3262 { 3263 int error = 0; 3264 nvlist_t *nvprops = NULL; 3265 char *origin_name; 3266 3267 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0) 3268 return (SET_ERROR(EINVAL)); 3269 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops); 3270 3271 if (strchr(fsname, '@') || 3272 strchr(fsname, '%')) 3273 return (SET_ERROR(EINVAL)); 3274 3275 if (dataset_namecheck(origin_name, NULL, NULL) != 0) 3276 return (SET_ERROR(EINVAL)); 3277 error = dmu_objset_clone(fsname, origin_name); 3278 if (error != 0) 3279 return (error); 3280 3281 /* 3282 * It would be nice to do this atomically. 3283 */ 3284 if (error == 0) { 3285 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL, 3286 nvprops, outnvl); 3287 if (error != 0) 3288 (void) dsl_destroy_head(fsname); 3289 } 3290 return (error); 3291 } 3292 3293 /* 3294 * innvl: { 3295 * "snaps" -> { snapshot1, snapshot2 } 3296 * (optional) "props" -> { prop -> value (string) } 3297 * } 3298 * 3299 * outnvl: snapshot -> error code (int32) 3300 */ 3301 static int 3302 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3303 { 3304 nvlist_t *snaps; 3305 nvlist_t *props = NULL; 3306 int error, poollen; 3307 nvpair_t *pair; 3308 3309 (void) nvlist_lookup_nvlist(innvl, "props", &props); 3310 if ((error = zfs_check_userprops(poolname, props)) != 0) 3311 return (error); 3312 3313 if (!nvlist_empty(props) && 3314 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS)) 3315 return (SET_ERROR(ENOTSUP)); 3316 3317 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3318 return (SET_ERROR(EINVAL)); 3319 poollen = strlen(poolname); 3320 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3321 pair = nvlist_next_nvpair(snaps, pair)) { 3322 const char *name = nvpair_name(pair); 3323 const char *cp = strchr(name, '@'); 3324 3325 /* 3326 * The snap name must contain an @, and the part after it must 3327 * contain only valid characters. 3328 */ 3329 if (cp == NULL || 3330 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3331 return (SET_ERROR(EINVAL)); 3332 3333 /* 3334 * The snap must be in the specified pool. 3335 */ 3336 if (strncmp(name, poolname, poollen) != 0 || 3337 (name[poollen] != '/' && name[poollen] != '@')) 3338 return (SET_ERROR(EXDEV)); 3339 3340 /* This must be the only snap of this fs. */ 3341 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair); 3342 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) { 3343 if (strncmp(name, nvpair_name(pair2), cp - name + 1) 3344 == 0) { 3345 return (SET_ERROR(EXDEV)); 3346 } 3347 } 3348 } 3349 3350 error = dsl_dataset_snapshot(snaps, props, outnvl); 3351 return (error); 3352 } 3353 3354 /* 3355 * innvl: "message" -> string 3356 */ 3357 /* ARGSUSED */ 3358 static int 3359 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl) 3360 { 3361 char *message; 3362 spa_t *spa; 3363 int error; 3364 char *poolname; 3365 3366 /* 3367 * The poolname in the ioctl is not set, we get it from the TSD, 3368 * which was set at the end of the last successful ioctl that allows 3369 * logging. The secpolicy func already checked that it is set. 3370 * Only one log ioctl is allowed after each successful ioctl, so 3371 * we clear the TSD here. 3372 */ 3373 poolname = tsd_get(zfs_allow_log_key); 3374 (void) tsd_set(zfs_allow_log_key, NULL); 3375 error = spa_open(poolname, &spa, FTAG); 3376 strfree(poolname); 3377 if (error != 0) 3378 return (error); 3379 3380 if (nvlist_lookup_string(innvl, "message", &message) != 0) { 3381 spa_close(spa, FTAG); 3382 return (SET_ERROR(EINVAL)); 3383 } 3384 3385 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) { 3386 spa_close(spa, FTAG); 3387 return (SET_ERROR(ENOTSUP)); 3388 } 3389 3390 error = spa_history_log(spa, message); 3391 spa_close(spa, FTAG); 3392 return (error); 3393 } 3394 3395 /* 3396 * The dp_config_rwlock must not be held when calling this, because the 3397 * unmount may need to write out data. 3398 * 3399 * This function is best-effort. Callers must deal gracefully if it 3400 * remains mounted (or is remounted after this call). 3401 * 3402 * Returns 0 if the argument is not a snapshot, or it is not currently a 3403 * filesystem, or we were able to unmount it. Returns error code otherwise. 3404 */ 3405 int 3406 zfs_unmount_snap(const char *snapname) 3407 { 3408 vfs_t *vfsp; 3409 zfsvfs_t *zfsvfs; 3410 int err; 3411 3412 if (strchr(snapname, '@') == NULL) 3413 return (0); 3414 3415 vfsp = zfs_get_vfs(snapname); 3416 if (vfsp == NULL) 3417 return (0); 3418 3419 zfsvfs = vfsp->vfs_data; 3420 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os))); 3421 3422 err = vn_vfswlock(vfsp->vfs_vnodecovered); 3423 VFS_RELE(vfsp); 3424 if (err != 0) 3425 return (SET_ERROR(err)); 3426 3427 /* 3428 * Always force the unmount for snapshots. 3429 */ 3430 (void) dounmount(vfsp, MS_FORCE, kcred); 3431 return (0); 3432 } 3433 3434 /* ARGSUSED */ 3435 static int 3436 zfs_unmount_snap_cb(const char *snapname, void *arg) 3437 { 3438 return (zfs_unmount_snap(snapname)); 3439 } 3440 3441 /* 3442 * When a clone is destroyed, its origin may also need to be destroyed, 3443 * in which case it must be unmounted. This routine will do that unmount 3444 * if necessary. 3445 */ 3446 void 3447 zfs_destroy_unmount_origin(const char *fsname) 3448 { 3449 int error; 3450 objset_t *os; 3451 dsl_dataset_t *ds; 3452 3453 error = dmu_objset_hold(fsname, FTAG, &os); 3454 if (error != 0) 3455 return; 3456 ds = dmu_objset_ds(os); 3457 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) { 3458 char originname[MAXNAMELEN]; 3459 dsl_dataset_name(ds->ds_prev, originname); 3460 dmu_objset_rele(os, FTAG); 3461 (void) zfs_unmount_snap(originname); 3462 } else { 3463 dmu_objset_rele(os, FTAG); 3464 } 3465 } 3466 3467 /* 3468 * innvl: { 3469 * "snaps" -> { snapshot1, snapshot2 } 3470 * (optional boolean) "defer" 3471 * } 3472 * 3473 * outnvl: snapshot -> error code (int32) 3474 * 3475 */ 3476 /* ARGSUSED */ 3477 static int 3478 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3479 { 3480 nvlist_t *snaps; 3481 nvpair_t *pair; 3482 boolean_t defer; 3483 3484 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0) 3485 return (SET_ERROR(EINVAL)); 3486 defer = nvlist_exists(innvl, "defer"); 3487 3488 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; 3489 pair = nvlist_next_nvpair(snaps, pair)) { 3490 (void) zfs_unmount_snap(nvpair_name(pair)); 3491 } 3492 3493 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl)); 3494 } 3495 3496 /* 3497 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>. 3498 * All bookmarks must be in the same pool. 3499 * 3500 * innvl: { 3501 * bookmark1 -> snapshot1, bookmark2 -> snapshot2 3502 * } 3503 * 3504 * outnvl: bookmark -> error code (int32) 3505 * 3506 */ 3507 /* ARGSUSED */ 3508 static int 3509 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) 3510 { 3511 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3512 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3513 char *snap_name; 3514 3515 /* 3516 * Verify the snapshot argument. 3517 */ 3518 if (nvpair_value_string(pair, &snap_name) != 0) 3519 return (SET_ERROR(EINVAL)); 3520 3521 3522 /* Verify that the keys (bookmarks) are unique */ 3523 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair); 3524 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) { 3525 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0) 3526 return (SET_ERROR(EINVAL)); 3527 } 3528 } 3529 3530 return (dsl_bookmark_create(innvl, outnvl)); 3531 } 3532 3533 /* 3534 * innvl: { 3535 * property 1, property 2, ... 3536 * } 3537 * 3538 * outnvl: { 3539 * bookmark name 1 -> { property 1, property 2, ... }, 3540 * bookmark name 2 -> { property 1, property 2, ... } 3541 * } 3542 * 3543 */ 3544 static int 3545 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) 3546 { 3547 return (dsl_get_bookmarks(fsname, innvl, outnvl)); 3548 } 3549 3550 /* 3551 * innvl: { 3552 * bookmark name 1, bookmark name 2 3553 * } 3554 * 3555 * outnvl: bookmark -> error code (int32) 3556 * 3557 */ 3558 static int 3559 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl, 3560 nvlist_t *outnvl) 3561 { 3562 int error, poollen; 3563 3564 poollen = strlen(poolname); 3565 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL); 3566 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) { 3567 const char *name = nvpair_name(pair); 3568 const char *cp = strchr(name, '#'); 3569 3570 /* 3571 * The bookmark name must contain an #, and the part after it 3572 * must contain only valid characters. 3573 */ 3574 if (cp == NULL || 3575 zfs_component_namecheck(cp + 1, NULL, NULL) != 0) 3576 return (SET_ERROR(EINVAL)); 3577 3578 /* 3579 * The bookmark must be in the specified pool. 3580 */ 3581 if (strncmp(name, poolname, poollen) != 0 || 3582 (name[poollen] != '/' && name[poollen] != '#')) 3583 return (SET_ERROR(EXDEV)); 3584 } 3585 3586 error = dsl_bookmark_destroy(innvl, outnvl); 3587 return (error); 3588 } 3589 3590 /* 3591 * inputs: 3592 * zc_name name of dataset to destroy 3593 * zc_objset_type type of objset 3594 * zc_defer_destroy mark for deferred destroy 3595 * 3596 * outputs: none 3597 */ 3598 static int 3599 zfs_ioc_destroy(zfs_cmd_t *zc) 3600 { 3601 int err; 3602 3603 if (zc->zc_objset_type == DMU_OST_ZFS) { 3604 err = zfs_unmount_snap(zc->zc_name); 3605 if (err != 0) 3606 return (err); 3607 } 3608 3609 if (strchr(zc->zc_name, '@')) 3610 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy); 3611 else 3612 err = dsl_destroy_head(zc->zc_name); 3613 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0) 3614 (void) zvol_remove_minor(zc->zc_name); 3615 return (err); 3616 } 3617 3618 /* 3619 * fsname is name of dataset to rollback (to most recent snapshot) 3620 * 3621 * innvl is not used. 3622 * 3623 * outnvl: "target" -> name of most recent snapshot 3624 * } 3625 */ 3626 /* ARGSUSED */ 3627 static int 3628 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl) 3629 { 3630 zfsvfs_t *zfsvfs; 3631 int error; 3632 3633 if (getzfsvfs(fsname, &zfsvfs) == 0) { 3634 error = zfs_suspend_fs(zfsvfs); 3635 if (error == 0) { 3636 int resume_err; 3637 3638 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl); 3639 resume_err = zfs_resume_fs(zfsvfs, fsname); 3640 error = error ? error : resume_err; 3641 } 3642 VFS_RELE(zfsvfs->z_vfs); 3643 } else { 3644 error = dsl_dataset_rollback(fsname, NULL, outnvl); 3645 } 3646 return (error); 3647 } 3648 3649 static int 3650 recursive_unmount(const char *fsname, void *arg) 3651 { 3652 const char *snapname = arg; 3653 char fullname[MAXNAMELEN]; 3654 3655 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname); 3656 return (zfs_unmount_snap(fullname)); 3657 } 3658 3659 /* 3660 * inputs: 3661 * zc_name old name of dataset 3662 * zc_value new name of dataset 3663 * zc_cookie recursive flag (only valid for snapshots) 3664 * 3665 * outputs: none 3666 */ 3667 static int 3668 zfs_ioc_rename(zfs_cmd_t *zc) 3669 { 3670 boolean_t recursive = zc->zc_cookie & 1; 3671 char *at; 3672 3673 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0'; 3674 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 3675 strchr(zc->zc_value, '%')) 3676 return (SET_ERROR(EINVAL)); 3677 3678 at = strchr(zc->zc_name, '@'); 3679 if (at != NULL) { 3680 /* snaps must be in same fs */ 3681 int error; 3682 3683 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1)) 3684 return (SET_ERROR(EXDEV)); 3685 *at = '\0'; 3686 if (zc->zc_objset_type == DMU_OST_ZFS) { 3687 error = dmu_objset_find(zc->zc_name, 3688 recursive_unmount, at + 1, 3689 recursive ? DS_FIND_CHILDREN : 0); 3690 if (error != 0) { 3691 *at = '@'; 3692 return (error); 3693 } 3694 } 3695 error = dsl_dataset_rename_snapshot(zc->zc_name, 3696 at + 1, strchr(zc->zc_value, '@') + 1, recursive); 3697 *at = '@'; 3698 3699 return (error); 3700 } else { 3701 if (zc->zc_objset_type == DMU_OST_ZVOL) 3702 (void) zvol_remove_minor(zc->zc_name); 3703 return (dsl_dir_rename(zc->zc_name, zc->zc_value)); 3704 } 3705 } 3706 3707 static int 3708 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) 3709 { 3710 const char *propname = nvpair_name(pair); 3711 boolean_t issnap = (strchr(dsname, '@') != NULL); 3712 zfs_prop_t prop = zfs_name_to_prop(propname); 3713 uint64_t intval; 3714 int err; 3715 3716 if (prop == ZPROP_INVAL) { 3717 if (zfs_prop_user(propname)) { 3718 if (err = zfs_secpolicy_write_perms(dsname, 3719 ZFS_DELEG_PERM_USERPROP, cr)) 3720 return (err); 3721 return (0); 3722 } 3723 3724 if (!issnap && zfs_prop_userquota(propname)) { 3725 const char *perm = NULL; 3726 const char *uq_prefix = 3727 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA]; 3728 const char *gq_prefix = 3729 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA]; 3730 3731 if (strncmp(propname, uq_prefix, 3732 strlen(uq_prefix)) == 0) { 3733 perm = ZFS_DELEG_PERM_USERQUOTA; 3734 } else if (strncmp(propname, gq_prefix, 3735 strlen(gq_prefix)) == 0) { 3736 perm = ZFS_DELEG_PERM_GROUPQUOTA; 3737 } else { 3738 /* USERUSED and GROUPUSED are read-only */ 3739 return (SET_ERROR(EINVAL)); 3740 } 3741 3742 if (err = zfs_secpolicy_write_perms(dsname, perm, cr)) 3743 return (err); 3744 return (0); 3745 } 3746 3747 return (SET_ERROR(EINVAL)); 3748 } 3749 3750 if (issnap) 3751 return (SET_ERROR(EINVAL)); 3752 3753 if (nvpair_type(pair) == DATA_TYPE_NVLIST) { 3754 /* 3755 * dsl_prop_get_all_impl() returns properties in this 3756 * format. 3757 */ 3758 nvlist_t *attrs; 3759 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); 3760 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3761 &pair) == 0); 3762 } 3763 3764 /* 3765 * Check that this value is valid for this pool version 3766 */ 3767 switch (prop) { 3768 case ZFS_PROP_COMPRESSION: 3769 /* 3770 * If the user specified gzip compression, make sure 3771 * the SPA supports it. We ignore any errors here since 3772 * we'll catch them later. 3773 */ 3774 if (nvpair_value_uint64(pair, &intval) == 0) { 3775 if (intval >= ZIO_COMPRESS_GZIP_1 && 3776 intval <= ZIO_COMPRESS_GZIP_9 && 3777 zfs_earlier_version(dsname, 3778 SPA_VERSION_GZIP_COMPRESSION)) { 3779 return (SET_ERROR(ENOTSUP)); 3780 } 3781 3782 if (intval == ZIO_COMPRESS_ZLE && 3783 zfs_earlier_version(dsname, 3784 SPA_VERSION_ZLE_COMPRESSION)) 3785 return (SET_ERROR(ENOTSUP)); 3786 3787 if (intval == ZIO_COMPRESS_LZ4) { 3788 spa_t *spa; 3789 3790 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3791 return (err); 3792 3793 if (!spa_feature_is_enabled(spa, 3794 SPA_FEATURE_LZ4_COMPRESS)) { 3795 spa_close(spa, FTAG); 3796 return (SET_ERROR(ENOTSUP)); 3797 } 3798 spa_close(spa, FTAG); 3799 } 3800 3801 /* 3802 * If this is a bootable dataset then 3803 * verify that the compression algorithm 3804 * is supported for booting. We must return 3805 * something other than ENOTSUP since it 3806 * implies a downrev pool version. 3807 */ 3808 if (zfs_is_bootfs(dsname) && 3809 !BOOTFS_COMPRESS_VALID(intval)) { 3810 return (SET_ERROR(ERANGE)); 3811 } 3812 } 3813 break; 3814 3815 case ZFS_PROP_COPIES: 3816 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS)) 3817 return (SET_ERROR(ENOTSUP)); 3818 break; 3819 3820 case ZFS_PROP_DEDUP: 3821 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP)) 3822 return (SET_ERROR(ENOTSUP)); 3823 break; 3824 3825 case ZFS_PROP_RECORDSIZE: 3826 /* Record sizes above 128k need the feature to be enabled */ 3827 if (nvpair_value_uint64(pair, &intval) == 0 && 3828 intval > SPA_OLD_MAXBLOCKSIZE) { 3829 spa_t *spa; 3830 3831 /* 3832 * If this is a bootable dataset then 3833 * the we don't allow large (>128K) blocks, 3834 * because GRUB doesn't support them. 3835 */ 3836 if (zfs_is_bootfs(dsname) && 3837 intval > SPA_OLD_MAXBLOCKSIZE) { 3838 return (SET_ERROR(EDOM)); 3839 } 3840 3841 /* 3842 * We don't allow setting the property above 1MB, 3843 * unless the tunable has been changed. 3844 */ 3845 if (intval > zfs_max_recordsize || 3846 intval > SPA_MAXBLOCKSIZE) 3847 return (SET_ERROR(EDOM)); 3848 3849 if ((err = spa_open(dsname, &spa, FTAG)) != 0) 3850 return (err); 3851 3852 if (!spa_feature_is_enabled(spa, 3853 SPA_FEATURE_LARGE_BLOCKS)) { 3854 spa_close(spa, FTAG); 3855 return (SET_ERROR(ENOTSUP)); 3856 } 3857 spa_close(spa, FTAG); 3858 } 3859 break; 3860 3861 case ZFS_PROP_SHARESMB: 3862 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) 3863 return (SET_ERROR(ENOTSUP)); 3864 break; 3865 3866 case ZFS_PROP_ACLINHERIT: 3867 if (nvpair_type(pair) == DATA_TYPE_UINT64 && 3868 nvpair_value_uint64(pair, &intval) == 0) { 3869 if (intval == ZFS_ACL_PASSTHROUGH_X && 3870 zfs_earlier_version(dsname, 3871 SPA_VERSION_PASSTHROUGH_X)) 3872 return (SET_ERROR(ENOTSUP)); 3873 } 3874 break; 3875 } 3876 3877 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED())); 3878 } 3879 3880 /* 3881 * Checks for a race condition to make sure we don't increment a feature flag 3882 * multiple times. 3883 */ 3884 static int 3885 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx) 3886 { 3887 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3888 spa_feature_t *featurep = arg; 3889 3890 if (!spa_feature_is_active(spa, *featurep)) 3891 return (0); 3892 else 3893 return (SET_ERROR(EBUSY)); 3894 } 3895 3896 /* 3897 * The callback invoked on feature activation in the sync task caused by 3898 * zfs_prop_activate_feature. 3899 */ 3900 static void 3901 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx) 3902 { 3903 spa_t *spa = dmu_tx_pool(tx)->dp_spa; 3904 spa_feature_t *featurep = arg; 3905 3906 spa_feature_incr(spa, *featurep, tx); 3907 } 3908 3909 /* 3910 * Activates a feature on a pool in response to a property setting. This 3911 * creates a new sync task which modifies the pool to reflect the feature 3912 * as being active. 3913 */ 3914 static int 3915 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature) 3916 { 3917 int err; 3918 3919 /* EBUSY here indicates that the feature is already active */ 3920 err = dsl_sync_task(spa_name(spa), 3921 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync, 3922 &feature, 2, ZFS_SPACE_CHECK_RESERVED); 3923 3924 if (err != 0 && err != EBUSY) 3925 return (err); 3926 else 3927 return (0); 3928 } 3929 3930 /* 3931 * Removes properties from the given props list that fail permission checks 3932 * needed to clear them and to restore them in case of a receive error. For each 3933 * property, make sure we have both set and inherit permissions. 3934 * 3935 * Returns the first error encountered if any permission checks fail. If the 3936 * caller provides a non-NULL errlist, it also gives the complete list of names 3937 * of all the properties that failed a permission check along with the 3938 * corresponding error numbers. The caller is responsible for freeing the 3939 * returned errlist. 3940 * 3941 * If every property checks out successfully, zero is returned and the list 3942 * pointed at by errlist is NULL. 3943 */ 3944 static int 3945 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist) 3946 { 3947 zfs_cmd_t *zc; 3948 nvpair_t *pair, *next_pair; 3949 nvlist_t *errors; 3950 int err, rv = 0; 3951 3952 if (props == NULL) 3953 return (0); 3954 3955 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3956 3957 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); 3958 (void) strcpy(zc->zc_name, dataset); 3959 pair = nvlist_next_nvpair(props, NULL); 3960 while (pair != NULL) { 3961 next_pair = nvlist_next_nvpair(props, pair); 3962 3963 (void) strcpy(zc->zc_value, nvpair_name(pair)); 3964 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || 3965 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { 3966 VERIFY(nvlist_remove_nvpair(props, pair) == 0); 3967 VERIFY(nvlist_add_int32(errors, 3968 zc->zc_value, err) == 0); 3969 } 3970 pair = next_pair; 3971 } 3972 kmem_free(zc, sizeof (zfs_cmd_t)); 3973 3974 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) { 3975 nvlist_free(errors); 3976 errors = NULL; 3977 } else { 3978 VERIFY(nvpair_value_int32(pair, &rv) == 0); 3979 } 3980 3981 if (errlist == NULL) 3982 nvlist_free(errors); 3983 else 3984 *errlist = errors; 3985 3986 return (rv); 3987 } 3988 3989 static boolean_t 3990 propval_equals(nvpair_t *p1, nvpair_t *p2) 3991 { 3992 if (nvpair_type(p1) == DATA_TYPE_NVLIST) { 3993 /* dsl_prop_get_all_impl() format */ 3994 nvlist_t *attrs; 3995 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); 3996 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 3997 &p1) == 0); 3998 } 3999 4000 if (nvpair_type(p2) == DATA_TYPE_NVLIST) { 4001 nvlist_t *attrs; 4002 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); 4003 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, 4004 &p2) == 0); 4005 } 4006 4007 if (nvpair_type(p1) != nvpair_type(p2)) 4008 return (B_FALSE); 4009 4010 if (nvpair_type(p1) == DATA_TYPE_STRING) { 4011 char *valstr1, *valstr2; 4012 4013 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0); 4014 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0); 4015 return (strcmp(valstr1, valstr2) == 0); 4016 } else { 4017 uint64_t intval1, intval2; 4018 4019 VERIFY(nvpair_value_uint64(p1, &intval1) == 0); 4020 VERIFY(nvpair_value_uint64(p2, &intval2) == 0); 4021 return (intval1 == intval2); 4022 } 4023 } 4024 4025 /* 4026 * Remove properties from props if they are not going to change (as determined 4027 * by comparison with origprops). Remove them from origprops as well, since we 4028 * do not need to clear or restore properties that won't change. 4029 */ 4030 static void 4031 props_reduce(nvlist_t *props, nvlist_t *origprops) 4032 { 4033 nvpair_t *pair, *next_pair; 4034 4035 if (origprops == NULL) 4036 return; /* all props need to be received */ 4037 4038 pair = nvlist_next_nvpair(props, NULL); 4039 while (pair != NULL) { 4040 const char *propname = nvpair_name(pair); 4041 nvpair_t *match; 4042 4043 next_pair = nvlist_next_nvpair(props, pair); 4044 4045 if ((nvlist_lookup_nvpair(origprops, propname, 4046 &match) != 0) || !propval_equals(pair, match)) 4047 goto next; /* need to set received value */ 4048 4049 /* don't clear the existing received value */ 4050 (void) nvlist_remove_nvpair(origprops, match); 4051 /* don't bother receiving the property */ 4052 (void) nvlist_remove_nvpair(props, pair); 4053 next: 4054 pair = next_pair; 4055 } 4056 } 4057 4058 #ifdef DEBUG 4059 static boolean_t zfs_ioc_recv_inject_err; 4060 #endif 4061 4062 /* 4063 * inputs: 4064 * zc_name name of containing filesystem 4065 * zc_nvlist_src{_size} nvlist of properties to apply 4066 * zc_value name of snapshot to create 4067 * zc_string name of clone origin (if DRR_FLAG_CLONE) 4068 * zc_cookie file descriptor to recv from 4069 * zc_begin_record the BEGIN record of the stream (not byteswapped) 4070 * zc_guid force flag 4071 * zc_cleanup_fd cleanup-on-exit file descriptor 4072 * zc_action_handle handle for this guid/ds mapping (or zero on first call) 4073 * 4074 * outputs: 4075 * zc_cookie number of bytes read 4076 * zc_nvlist_dst{_size} error for each unapplied received property 4077 * zc_obj zprop_errflags_t 4078 * zc_action_handle handle for this guid/ds mapping 4079 */ 4080 static int 4081 zfs_ioc_recv(zfs_cmd_t *zc) 4082 { 4083 file_t *fp; 4084 dmu_recv_cookie_t drc; 4085 boolean_t force = (boolean_t)zc->zc_guid; 4086 int fd; 4087 int error = 0; 4088 int props_error = 0; 4089 nvlist_t *errors; 4090 offset_t off; 4091 nvlist_t *props = NULL; /* sent properties */ 4092 nvlist_t *origprops = NULL; /* existing properties */ 4093 char *origin = NULL; 4094 char *tosnap; 4095 char tofs[ZFS_MAXNAMELEN]; 4096 boolean_t first_recvd_props = B_FALSE; 4097 4098 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 || 4099 strchr(zc->zc_value, '@') == NULL || 4100 strchr(zc->zc_value, '%')) 4101 return (SET_ERROR(EINVAL)); 4102 4103 (void) strcpy(tofs, zc->zc_value); 4104 tosnap = strchr(tofs, '@'); 4105 *tosnap++ = '\0'; 4106 4107 if (zc->zc_nvlist_src != NULL && 4108 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 4109 zc->zc_iflags, &props)) != 0) 4110 return (error); 4111 4112 fd = zc->zc_cookie; 4113 fp = getf(fd); 4114 if (fp == NULL) { 4115 nvlist_free(props); 4116 return (SET_ERROR(EBADF)); 4117 } 4118 4119 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); 4120 4121 if (zc->zc_string[0]) 4122 origin = zc->zc_string; 4123 4124 error = dmu_recv_begin(tofs, tosnap, 4125 &zc->zc_begin_record, force, origin, &drc); 4126 if (error != 0) 4127 goto out; 4128 4129 /* 4130 * Set properties before we receive the stream so that they are applied 4131 * to the new data. Note that we must call dmu_recv_stream() if 4132 * dmu_recv_begin() succeeds. 4133 */ 4134 if (props != NULL && !drc.drc_newfs) { 4135 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >= 4136 SPA_VERSION_RECVD_PROPS && 4137 !dsl_prop_get_hasrecvd(tofs)) 4138 first_recvd_props = B_TRUE; 4139 4140 /* 4141 * If new received properties are supplied, they are to 4142 * completely replace the existing received properties, so stash 4143 * away the existing ones. 4144 */ 4145 if (dsl_prop_get_received(tofs, &origprops) == 0) { 4146 nvlist_t *errlist = NULL; 4147 /* 4148 * Don't bother writing a property if its value won't 4149 * change (and avoid the unnecessary security checks). 4150 * 4151 * The first receive after SPA_VERSION_RECVD_PROPS is a 4152 * special case where we blow away all local properties 4153 * regardless. 4154 */ 4155 if (!first_recvd_props) 4156 props_reduce(props, origprops); 4157 if (zfs_check_clearable(tofs, origprops, &errlist) != 0) 4158 (void) nvlist_merge(errors, errlist, 0); 4159 nvlist_free(errlist); 4160 4161 if (clear_received_props(tofs, origprops, 4162 first_recvd_props ? NULL : props) != 0) 4163 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4164 } else { 4165 zc->zc_obj |= ZPROP_ERR_NOCLEAR; 4166 } 4167 } 4168 4169 if (props != NULL) { 4170 props_error = dsl_prop_set_hasrecvd(tofs); 4171 4172 if (props_error == 0) { 4173 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED, 4174 props, errors); 4175 } 4176 } 4177 4178 if (zc->zc_nvlist_dst_size != 0 && 4179 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 || 4180 put_nvlist(zc, errors) != 0)) { 4181 /* 4182 * Caller made zc->zc_nvlist_dst less than the minimum expected 4183 * size or supplied an invalid address. 4184 */ 4185 props_error = SET_ERROR(EINVAL); 4186 } 4187 4188 off = fp->f_offset; 4189 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd, 4190 &zc->zc_action_handle); 4191 4192 if (error == 0) { 4193 zfsvfs_t *zfsvfs = NULL; 4194 4195 if (getzfsvfs(tofs, &zfsvfs) == 0) { 4196 /* online recv */ 4197 int end_err; 4198 4199 error = zfs_suspend_fs(zfsvfs); 4200 /* 4201 * If the suspend fails, then the recv_end will 4202 * likely also fail, and clean up after itself. 4203 */ 4204 end_err = dmu_recv_end(&drc, zfsvfs); 4205 if (error == 0) 4206 error = zfs_resume_fs(zfsvfs, tofs); 4207 error = error ? error : end_err; 4208 VFS_RELE(zfsvfs->z_vfs); 4209 } else { 4210 error = dmu_recv_end(&drc, NULL); 4211 } 4212 } 4213 4214 zc->zc_cookie = off - fp->f_offset; 4215 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 4216 fp->f_offset = off; 4217 4218 #ifdef DEBUG 4219 if (zfs_ioc_recv_inject_err) { 4220 zfs_ioc_recv_inject_err = B_FALSE; 4221 error = 1; 4222 } 4223 #endif 4224 /* 4225 * On error, restore the original props. 4226 */ 4227 if (error != 0 && props != NULL && !drc.drc_newfs) { 4228 if (clear_received_props(tofs, props, NULL) != 0) { 4229 /* 4230 * We failed to clear the received properties. 4231 * Since we may have left a $recvd value on the 4232 * system, we can't clear the $hasrecvd flag. 4233 */ 4234 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4235 } else if (first_recvd_props) { 4236 dsl_prop_unset_hasrecvd(tofs); 4237 } 4238 4239 if (origprops == NULL && !drc.drc_newfs) { 4240 /* We failed to stash the original properties. */ 4241 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4242 } 4243 4244 /* 4245 * dsl_props_set() will not convert RECEIVED to LOCAL on or 4246 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL 4247 * explictly if we're restoring local properties cleared in the 4248 * first new-style receive. 4249 */ 4250 if (origprops != NULL && 4251 zfs_set_prop_nvlist(tofs, (first_recvd_props ? 4252 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED), 4253 origprops, NULL) != 0) { 4254 /* 4255 * We stashed the original properties but failed to 4256 * restore them. 4257 */ 4258 zc->zc_obj |= ZPROP_ERR_NORESTORE; 4259 } 4260 } 4261 out: 4262 nvlist_free(props); 4263 nvlist_free(origprops); 4264 nvlist_free(errors); 4265 releasef(fd); 4266 4267 if (error == 0) 4268 error = props_error; 4269 4270 return (error); 4271 } 4272 4273 /* 4274 * inputs: 4275 * zc_name name of snapshot to send 4276 * zc_cookie file descriptor to send stream to 4277 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj) 4278 * zc_sendobj objsetid of snapshot to send 4279 * zc_fromobj objsetid of incremental fromsnap (may be zero) 4280 * zc_guid if set, estimate size of stream only. zc_cookie is ignored. 4281 * output size in zc_objset_type. 4282 * zc_flags lzc_send_flags 4283 * 4284 * outputs: 4285 * zc_objset_type estimated size, if zc_guid is set 4286 */ 4287 static int 4288 zfs_ioc_send(zfs_cmd_t *zc) 4289 { 4290 int error; 4291 offset_t off; 4292 boolean_t estimate = (zc->zc_guid != 0); 4293 boolean_t embedok = (zc->zc_flags & 0x1); 4294 boolean_t large_block_ok = (zc->zc_flags & 0x2); 4295 4296 if (zc->zc_obj != 0) { 4297 dsl_pool_t *dp; 4298 dsl_dataset_t *tosnap; 4299 4300 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4301 if (error != 0) 4302 return (error); 4303 4304 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4305 if (error != 0) { 4306 dsl_pool_rele(dp, FTAG); 4307 return (error); 4308 } 4309 4310 if (dsl_dir_is_clone(tosnap->ds_dir)) 4311 zc->zc_fromobj = 4312 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj; 4313 dsl_dataset_rele(tosnap, FTAG); 4314 dsl_pool_rele(dp, FTAG); 4315 } 4316 4317 if (estimate) { 4318 dsl_pool_t *dp; 4319 dsl_dataset_t *tosnap; 4320 dsl_dataset_t *fromsnap = NULL; 4321 4322 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4323 if (error != 0) 4324 return (error); 4325 4326 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap); 4327 if (error != 0) { 4328 dsl_pool_rele(dp, FTAG); 4329 return (error); 4330 } 4331 4332 if (zc->zc_fromobj != 0) { 4333 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, 4334 FTAG, &fromsnap); 4335 if (error != 0) { 4336 dsl_dataset_rele(tosnap, FTAG); 4337 dsl_pool_rele(dp, FTAG); 4338 return (error); 4339 } 4340 } 4341 4342 error = dmu_send_estimate(tosnap, fromsnap, 4343 &zc->zc_objset_type); 4344 4345 if (fromsnap != NULL) 4346 dsl_dataset_rele(fromsnap, FTAG); 4347 dsl_dataset_rele(tosnap, FTAG); 4348 dsl_pool_rele(dp, FTAG); 4349 } else { 4350 file_t *fp = getf(zc->zc_cookie); 4351 if (fp == NULL) 4352 return (SET_ERROR(EBADF)); 4353 4354 off = fp->f_offset; 4355 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj, 4356 zc->zc_fromobj, embedok, large_block_ok, 4357 zc->zc_cookie, fp->f_vnode, &off); 4358 4359 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 4360 fp->f_offset = off; 4361 releasef(zc->zc_cookie); 4362 } 4363 return (error); 4364 } 4365 4366 /* 4367 * inputs: 4368 * zc_name name of snapshot on which to report progress 4369 * zc_cookie file descriptor of send stream 4370 * 4371 * outputs: 4372 * zc_cookie number of bytes written in send stream thus far 4373 */ 4374 static int 4375 zfs_ioc_send_progress(zfs_cmd_t *zc) 4376 { 4377 dsl_pool_t *dp; 4378 dsl_dataset_t *ds; 4379 dmu_sendarg_t *dsp = NULL; 4380 int error; 4381 4382 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 4383 if (error != 0) 4384 return (error); 4385 4386 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds); 4387 if (error != 0) { 4388 dsl_pool_rele(dp, FTAG); 4389 return (error); 4390 } 4391 4392 mutex_enter(&ds->ds_sendstream_lock); 4393 4394 /* 4395 * Iterate over all the send streams currently active on this dataset. 4396 * If there's one which matches the specified file descriptor _and_ the 4397 * stream was started by the current process, return the progress of 4398 * that stream. 4399 */ 4400 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL; 4401 dsp = list_next(&ds->ds_sendstreams, dsp)) { 4402 if (dsp->dsa_outfd == zc->zc_cookie && 4403 dsp->dsa_proc == curproc) 4404 break; 4405 } 4406 4407 if (dsp != NULL) 4408 zc->zc_cookie = *(dsp->dsa_off); 4409 else 4410 error = SET_ERROR(ENOENT); 4411 4412 mutex_exit(&ds->ds_sendstream_lock); 4413 dsl_dataset_rele(ds, FTAG); 4414 dsl_pool_rele(dp, FTAG); 4415 return (error); 4416 } 4417 4418 static int 4419 zfs_ioc_inject_fault(zfs_cmd_t *zc) 4420 { 4421 int id, error; 4422 4423 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id, 4424 &zc->zc_inject_record); 4425 4426 if (error == 0) 4427 zc->zc_guid = (uint64_t)id; 4428 4429 return (error); 4430 } 4431 4432 static int 4433 zfs_ioc_clear_fault(zfs_cmd_t *zc) 4434 { 4435 return (zio_clear_fault((int)zc->zc_guid)); 4436 } 4437 4438 static int 4439 zfs_ioc_inject_list_next(zfs_cmd_t *zc) 4440 { 4441 int id = (int)zc->zc_guid; 4442 int error; 4443 4444 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name), 4445 &zc->zc_inject_record); 4446 4447 zc->zc_guid = id; 4448 4449 return (error); 4450 } 4451 4452 static int 4453 zfs_ioc_error_log(zfs_cmd_t *zc) 4454 { 4455 spa_t *spa; 4456 int error; 4457 size_t count = (size_t)zc->zc_nvlist_dst_size; 4458 4459 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) 4460 return (error); 4461 4462 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst, 4463 &count); 4464 if (error == 0) 4465 zc->zc_nvlist_dst_size = count; 4466 else 4467 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa); 4468 4469 spa_close(spa, FTAG); 4470 4471 return (error); 4472 } 4473 4474 static int 4475 zfs_ioc_clear(zfs_cmd_t *zc) 4476 { 4477 spa_t *spa; 4478 vdev_t *vd; 4479 int error; 4480 4481 /* 4482 * On zpool clear we also fix up missing slogs 4483 */ 4484 mutex_enter(&spa_namespace_lock); 4485 spa = spa_lookup(zc->zc_name); 4486 if (spa == NULL) { 4487 mutex_exit(&spa_namespace_lock); 4488 return (SET_ERROR(EIO)); 4489 } 4490 if (spa_get_log_state(spa) == SPA_LOG_MISSING) { 4491 /* we need to let spa_open/spa_load clear the chains */ 4492 spa_set_log_state(spa, SPA_LOG_CLEAR); 4493 } 4494 spa->spa_last_open_failed = 0; 4495 mutex_exit(&spa_namespace_lock); 4496 4497 if (zc->zc_cookie & ZPOOL_NO_REWIND) { 4498 error = spa_open(zc->zc_name, &spa, FTAG); 4499 } else { 4500 nvlist_t *policy; 4501 nvlist_t *config = NULL; 4502 4503 if (zc->zc_nvlist_src == NULL) 4504 return (SET_ERROR(EINVAL)); 4505 4506 if ((error = get_nvlist(zc->zc_nvlist_src, 4507 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) { 4508 error = spa_open_rewind(zc->zc_name, &spa, FTAG, 4509 policy, &config); 4510 if (config != NULL) { 4511 int err; 4512 4513 if ((err = put_nvlist(zc, config)) != 0) 4514 error = err; 4515 nvlist_free(config); 4516 } 4517 nvlist_free(policy); 4518 } 4519 } 4520 4521 if (error != 0) 4522 return (error); 4523 4524 spa_vdev_state_enter(spa, SCL_NONE); 4525 4526 if (zc->zc_guid == 0) { 4527 vd = NULL; 4528 } else { 4529 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE); 4530 if (vd == NULL) { 4531 (void) spa_vdev_state_exit(spa, NULL, ENODEV); 4532 spa_close(spa, FTAG); 4533 return (SET_ERROR(ENODEV)); 4534 } 4535 } 4536 4537 vdev_clear(spa, vd); 4538 4539 (void) spa_vdev_state_exit(spa, NULL, 0); 4540 4541 /* 4542 * Resume any suspended I/Os. 4543 */ 4544 if (zio_resume(spa) != 0) 4545 error = SET_ERROR(EIO); 4546 4547 spa_close(spa, FTAG); 4548 4549 return (error); 4550 } 4551 4552 static int 4553 zfs_ioc_pool_reopen(zfs_cmd_t *zc) 4554 { 4555 spa_t *spa; 4556 int error; 4557 4558 error = spa_open(zc->zc_name, &spa, FTAG); 4559 if (error != 0) 4560 return (error); 4561 4562 spa_vdev_state_enter(spa, SCL_NONE); 4563 4564 /* 4565 * If a resilver is already in progress then set the 4566 * spa_scrub_reopen flag to B_TRUE so that we don't restart 4567 * the scan as a side effect of the reopen. Otherwise, let 4568 * vdev_open() decided if a resilver is required. 4569 */ 4570 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool); 4571 vdev_reopen(spa->spa_root_vdev); 4572 spa->spa_scrub_reopen = B_FALSE; 4573 4574 (void) spa_vdev_state_exit(spa, NULL, 0); 4575 spa_close(spa, FTAG); 4576 return (0); 4577 } 4578 /* 4579 * inputs: 4580 * zc_name name of filesystem 4581 * zc_value name of origin snapshot 4582 * 4583 * outputs: 4584 * zc_string name of conflicting snapshot, if there is one 4585 */ 4586 static int 4587 zfs_ioc_promote(zfs_cmd_t *zc) 4588 { 4589 char *cp; 4590 4591 /* 4592 * We don't need to unmount *all* the origin fs's snapshots, but 4593 * it's easier. 4594 */ 4595 cp = strchr(zc->zc_value, '@'); 4596 if (cp) 4597 *cp = '\0'; 4598 (void) dmu_objset_find(zc->zc_value, 4599 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS); 4600 return (dsl_dataset_promote(zc->zc_name, zc->zc_string)); 4601 } 4602 4603 /* 4604 * Retrieve a single {user|group}{used|quota}@... property. 4605 * 4606 * inputs: 4607 * zc_name name of filesystem 4608 * zc_objset_type zfs_userquota_prop_t 4609 * zc_value domain name (eg. "S-1-234-567-89") 4610 * zc_guid RID/UID/GID 4611 * 4612 * outputs: 4613 * zc_cookie property value 4614 */ 4615 static int 4616 zfs_ioc_userspace_one(zfs_cmd_t *zc) 4617 { 4618 zfsvfs_t *zfsvfs; 4619 int error; 4620 4621 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS) 4622 return (SET_ERROR(EINVAL)); 4623 4624 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4625 if (error != 0) 4626 return (error); 4627 4628 error = zfs_userspace_one(zfsvfs, 4629 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie); 4630 zfsvfs_rele(zfsvfs, FTAG); 4631 4632 return (error); 4633 } 4634 4635 /* 4636 * inputs: 4637 * zc_name name of filesystem 4638 * zc_cookie zap cursor 4639 * zc_objset_type zfs_userquota_prop_t 4640 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist) 4641 * 4642 * outputs: 4643 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t) 4644 * zc_cookie zap cursor 4645 */ 4646 static int 4647 zfs_ioc_userspace_many(zfs_cmd_t *zc) 4648 { 4649 zfsvfs_t *zfsvfs; 4650 int bufsize = zc->zc_nvlist_dst_size; 4651 4652 if (bufsize <= 0) 4653 return (SET_ERROR(ENOMEM)); 4654 4655 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE); 4656 if (error != 0) 4657 return (error); 4658 4659 void *buf = kmem_alloc(bufsize, KM_SLEEP); 4660 4661 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie, 4662 buf, &zc->zc_nvlist_dst_size); 4663 4664 if (error == 0) { 4665 error = xcopyout(buf, 4666 (void *)(uintptr_t)zc->zc_nvlist_dst, 4667 zc->zc_nvlist_dst_size); 4668 } 4669 kmem_free(buf, bufsize); 4670 zfsvfs_rele(zfsvfs, FTAG); 4671 4672 return (error); 4673 } 4674 4675 /* 4676 * inputs: 4677 * zc_name name of filesystem 4678 * 4679 * outputs: 4680 * none 4681 */ 4682 static int 4683 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc) 4684 { 4685 objset_t *os; 4686 int error = 0; 4687 zfsvfs_t *zfsvfs; 4688 4689 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) { 4690 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) { 4691 /* 4692 * If userused is not enabled, it may be because the 4693 * objset needs to be closed & reopened (to grow the 4694 * objset_phys_t). Suspend/resume the fs will do that. 4695 */ 4696 error = zfs_suspend_fs(zfsvfs); 4697 if (error == 0) { 4698 dmu_objset_refresh_ownership(zfsvfs->z_os, 4699 zfsvfs); 4700 error = zfs_resume_fs(zfsvfs, zc->zc_name); 4701 } 4702 } 4703 if (error == 0) 4704 error = dmu_objset_userspace_upgrade(zfsvfs->z_os); 4705 VFS_RELE(zfsvfs->z_vfs); 4706 } else { 4707 /* XXX kind of reading contents without owning */ 4708 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4709 if (error != 0) 4710 return (error); 4711 4712 error = dmu_objset_userspace_upgrade(os); 4713 dmu_objset_rele(os, FTAG); 4714 } 4715 4716 return (error); 4717 } 4718 4719 /* 4720 * We don't want to have a hard dependency 4721 * against some special symbols in sharefs 4722 * nfs, and smbsrv. Determine them if needed when 4723 * the first file system is shared. 4724 * Neither sharefs, nfs or smbsrv are unloadable modules. 4725 */ 4726 int (*znfsexport_fs)(void *arg); 4727 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t); 4728 int (*zsmbexport_fs)(void *arg, boolean_t add_share); 4729 4730 int zfs_nfsshare_inited; 4731 int zfs_smbshare_inited; 4732 4733 ddi_modhandle_t nfs_mod; 4734 ddi_modhandle_t sharefs_mod; 4735 ddi_modhandle_t smbsrv_mod; 4736 kmutex_t zfs_share_lock; 4737 4738 static int 4739 zfs_init_sharefs() 4740 { 4741 int error; 4742 4743 ASSERT(MUTEX_HELD(&zfs_share_lock)); 4744 /* Both NFS and SMB shares also require sharetab support. */ 4745 if (sharefs_mod == NULL && ((sharefs_mod = 4746 ddi_modopen("fs/sharefs", 4747 KRTLD_MODE_FIRST, &error)) == NULL)) { 4748 return (SET_ERROR(ENOSYS)); 4749 } 4750 if (zshare_fs == NULL && ((zshare_fs = 4751 (int (*)(enum sharefs_sys_op, share_t *, uint32_t)) 4752 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) { 4753 return (SET_ERROR(ENOSYS)); 4754 } 4755 return (0); 4756 } 4757 4758 static int 4759 zfs_ioc_share(zfs_cmd_t *zc) 4760 { 4761 int error; 4762 int opcode; 4763 4764 switch (zc->zc_share.z_sharetype) { 4765 case ZFS_SHARE_NFS: 4766 case ZFS_UNSHARE_NFS: 4767 if (zfs_nfsshare_inited == 0) { 4768 mutex_enter(&zfs_share_lock); 4769 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs", 4770 KRTLD_MODE_FIRST, &error)) == NULL)) { 4771 mutex_exit(&zfs_share_lock); 4772 return (SET_ERROR(ENOSYS)); 4773 } 4774 if (znfsexport_fs == NULL && 4775 ((znfsexport_fs = (int (*)(void *)) 4776 ddi_modsym(nfs_mod, 4777 "nfs_export", &error)) == NULL)) { 4778 mutex_exit(&zfs_share_lock); 4779 return (SET_ERROR(ENOSYS)); 4780 } 4781 error = zfs_init_sharefs(); 4782 if (error != 0) { 4783 mutex_exit(&zfs_share_lock); 4784 return (SET_ERROR(ENOSYS)); 4785 } 4786 zfs_nfsshare_inited = 1; 4787 mutex_exit(&zfs_share_lock); 4788 } 4789 break; 4790 case ZFS_SHARE_SMB: 4791 case ZFS_UNSHARE_SMB: 4792 if (zfs_smbshare_inited == 0) { 4793 mutex_enter(&zfs_share_lock); 4794 if (smbsrv_mod == NULL && ((smbsrv_mod = 4795 ddi_modopen("drv/smbsrv", 4796 KRTLD_MODE_FIRST, &error)) == NULL)) { 4797 mutex_exit(&zfs_share_lock); 4798 return (SET_ERROR(ENOSYS)); 4799 } 4800 if (zsmbexport_fs == NULL && ((zsmbexport_fs = 4801 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod, 4802 "smb_server_share", &error)) == NULL)) { 4803 mutex_exit(&zfs_share_lock); 4804 return (SET_ERROR(ENOSYS)); 4805 } 4806 error = zfs_init_sharefs(); 4807 if (error != 0) { 4808 mutex_exit(&zfs_share_lock); 4809 return (SET_ERROR(ENOSYS)); 4810 } 4811 zfs_smbshare_inited = 1; 4812 mutex_exit(&zfs_share_lock); 4813 } 4814 break; 4815 default: 4816 return (SET_ERROR(EINVAL)); 4817 } 4818 4819 switch (zc->zc_share.z_sharetype) { 4820 case ZFS_SHARE_NFS: 4821 case ZFS_UNSHARE_NFS: 4822 if (error = 4823 znfsexport_fs((void *) 4824 (uintptr_t)zc->zc_share.z_exportdata)) 4825 return (error); 4826 break; 4827 case ZFS_SHARE_SMB: 4828 case ZFS_UNSHARE_SMB: 4829 if (error = zsmbexport_fs((void *) 4830 (uintptr_t)zc->zc_share.z_exportdata, 4831 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ? 4832 B_TRUE: B_FALSE)) { 4833 return (error); 4834 } 4835 break; 4836 } 4837 4838 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS || 4839 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ? 4840 SHAREFS_ADD : SHAREFS_REMOVE; 4841 4842 /* 4843 * Add or remove share from sharetab 4844 */ 4845 error = zshare_fs(opcode, 4846 (void *)(uintptr_t)zc->zc_share.z_sharedata, 4847 zc->zc_share.z_sharemax); 4848 4849 return (error); 4850 4851 } 4852 4853 ace_t full_access[] = { 4854 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0} 4855 }; 4856 4857 /* 4858 * inputs: 4859 * zc_name name of containing filesystem 4860 * zc_obj object # beyond which we want next in-use object # 4861 * 4862 * outputs: 4863 * zc_obj next in-use object # 4864 */ 4865 static int 4866 zfs_ioc_next_obj(zfs_cmd_t *zc) 4867 { 4868 objset_t *os = NULL; 4869 int error; 4870 4871 error = dmu_objset_hold(zc->zc_name, FTAG, &os); 4872 if (error != 0) 4873 return (error); 4874 4875 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 4876 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg); 4877 4878 dmu_objset_rele(os, FTAG); 4879 return (error); 4880 } 4881 4882 /* 4883 * inputs: 4884 * zc_name name of filesystem 4885 * zc_value prefix name for snapshot 4886 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process 4887 * 4888 * outputs: 4889 * zc_value short name of new snapshot 4890 */ 4891 static int 4892 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc) 4893 { 4894 char *snap_name; 4895 char *hold_name; 4896 int error; 4897 minor_t minor; 4898 4899 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor); 4900 if (error != 0) 4901 return (error); 4902 4903 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value, 4904 (u_longlong_t)ddi_get_lbolt64()); 4905 hold_name = kmem_asprintf("%%%s", zc->zc_value); 4906 4907 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor, 4908 hold_name); 4909 if (error == 0) 4910 (void) strcpy(zc->zc_value, snap_name); 4911 strfree(snap_name); 4912 strfree(hold_name); 4913 zfs_onexit_fd_rele(zc->zc_cleanup_fd); 4914 return (error); 4915 } 4916 4917 /* 4918 * inputs: 4919 * zc_name name of "to" snapshot 4920 * zc_value name of "from" snapshot 4921 * zc_cookie file descriptor to write diff data on 4922 * 4923 * outputs: 4924 * dmu_diff_record_t's to the file descriptor 4925 */ 4926 static int 4927 zfs_ioc_diff(zfs_cmd_t *zc) 4928 { 4929 file_t *fp; 4930 offset_t off; 4931 int error; 4932 4933 fp = getf(zc->zc_cookie); 4934 if (fp == NULL) 4935 return (SET_ERROR(EBADF)); 4936 4937 off = fp->f_offset; 4938 4939 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off); 4940 4941 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 4942 fp->f_offset = off; 4943 releasef(zc->zc_cookie); 4944 4945 return (error); 4946 } 4947 4948 /* 4949 * Remove all ACL files in shares dir 4950 */ 4951 static int 4952 zfs_smb_acl_purge(znode_t *dzp) 4953 { 4954 zap_cursor_t zc; 4955 zap_attribute_t zap; 4956 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 4957 int error; 4958 4959 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 4960 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 4961 zap_cursor_advance(&zc)) { 4962 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred, 4963 NULL, 0)) != 0) 4964 break; 4965 } 4966 zap_cursor_fini(&zc); 4967 return (error); 4968 } 4969 4970 static int 4971 zfs_ioc_smb_acl(zfs_cmd_t *zc) 4972 { 4973 vnode_t *vp; 4974 znode_t *dzp; 4975 vnode_t *resourcevp = NULL; 4976 znode_t *sharedir; 4977 zfsvfs_t *zfsvfs; 4978 nvlist_t *nvlist; 4979 char *src, *target; 4980 vattr_t vattr; 4981 vsecattr_t vsec; 4982 int error = 0; 4983 4984 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE, 4985 NO_FOLLOW, NULL, &vp)) != 0) 4986 return (error); 4987 4988 /* Now make sure mntpnt and dataset are ZFS */ 4989 4990 if (vp->v_vfsp->vfs_fstype != zfsfstype || 4991 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource), 4992 zc->zc_name) != 0)) { 4993 VN_RELE(vp); 4994 return (SET_ERROR(EINVAL)); 4995 } 4996 4997 dzp = VTOZ(vp); 4998 zfsvfs = dzp->z_zfsvfs; 4999 ZFS_ENTER(zfsvfs); 5000 5001 /* 5002 * Create share dir if its missing. 5003 */ 5004 mutex_enter(&zfsvfs->z_lock); 5005 if (zfsvfs->z_shares_dir == 0) { 5006 dmu_tx_t *tx; 5007 5008 tx = dmu_tx_create(zfsvfs->z_os); 5009 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE, 5010 ZFS_SHARES_DIR); 5011 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 5012 error = dmu_tx_assign(tx, TXG_WAIT); 5013 if (error != 0) { 5014 dmu_tx_abort(tx); 5015 } else { 5016 error = zfs_create_share_dir(zfsvfs, tx); 5017 dmu_tx_commit(tx); 5018 } 5019 if (error != 0) { 5020 mutex_exit(&zfsvfs->z_lock); 5021 VN_RELE(vp); 5022 ZFS_EXIT(zfsvfs); 5023 return (error); 5024 } 5025 } 5026 mutex_exit(&zfsvfs->z_lock); 5027 5028 ASSERT(zfsvfs->z_shares_dir); 5029 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) { 5030 VN_RELE(vp); 5031 ZFS_EXIT(zfsvfs); 5032 return (error); 5033 } 5034 5035 switch (zc->zc_cookie) { 5036 case ZFS_SMB_ACL_ADD: 5037 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 5038 vattr.va_type = VREG; 5039 vattr.va_mode = S_IFREG|0777; 5040 vattr.va_uid = 0; 5041 vattr.va_gid = 0; 5042 5043 vsec.vsa_mask = VSA_ACE; 5044 vsec.vsa_aclentp = &full_access; 5045 vsec.vsa_aclentsz = sizeof (full_access); 5046 vsec.vsa_aclcnt = 1; 5047 5048 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string, 5049 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec); 5050 if (resourcevp) 5051 VN_RELE(resourcevp); 5052 break; 5053 5054 case ZFS_SMB_ACL_REMOVE: 5055 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred, 5056 NULL, 0); 5057 break; 5058 5059 case ZFS_SMB_ACL_RENAME: 5060 if ((error = get_nvlist(zc->zc_nvlist_src, 5061 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) { 5062 VN_RELE(vp); 5063 ZFS_EXIT(zfsvfs); 5064 return (error); 5065 } 5066 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) || 5067 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET, 5068 &target)) { 5069 VN_RELE(vp); 5070 VN_RELE(ZTOV(sharedir)); 5071 ZFS_EXIT(zfsvfs); 5072 nvlist_free(nvlist); 5073 return (error); 5074 } 5075 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target, 5076 kcred, NULL, 0); 5077 nvlist_free(nvlist); 5078 break; 5079 5080 case ZFS_SMB_ACL_PURGE: 5081 error = zfs_smb_acl_purge(sharedir); 5082 break; 5083 5084 default: 5085 error = SET_ERROR(EINVAL); 5086 break; 5087 } 5088 5089 VN_RELE(vp); 5090 VN_RELE(ZTOV(sharedir)); 5091 5092 ZFS_EXIT(zfsvfs); 5093 5094 return (error); 5095 } 5096 5097 /* 5098 * innvl: { 5099 * "holds" -> { snapname -> holdname (string), ... } 5100 * (optional) "cleanup_fd" -> fd (int32) 5101 * } 5102 * 5103 * outnvl: { 5104 * snapname -> error value (int32) 5105 * ... 5106 * } 5107 */ 5108 /* ARGSUSED */ 5109 static int 5110 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist) 5111 { 5112 nvpair_t *pair; 5113 nvlist_t *holds; 5114 int cleanup_fd = -1; 5115 int error; 5116 minor_t minor = 0; 5117 5118 error = nvlist_lookup_nvlist(args, "holds", &holds); 5119 if (error != 0) 5120 return (SET_ERROR(EINVAL)); 5121 5122 /* make sure the user didn't pass us any invalid (empty) tags */ 5123 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL; 5124 pair = nvlist_next_nvpair(holds, pair)) { 5125 char *htag; 5126 5127 error = nvpair_value_string(pair, &htag); 5128 if (error != 0) 5129 return (SET_ERROR(error)); 5130 5131 if (strlen(htag) == 0) 5132 return (SET_ERROR(EINVAL)); 5133 } 5134 5135 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) { 5136 error = zfs_onexit_fd_hold(cleanup_fd, &minor); 5137 if (error != 0) 5138 return (error); 5139 } 5140 5141 error = dsl_dataset_user_hold(holds, minor, errlist); 5142 if (minor != 0) 5143 zfs_onexit_fd_rele(cleanup_fd); 5144 return (error); 5145 } 5146 5147 /* 5148 * innvl is not used. 5149 * 5150 * outnvl: { 5151 * holdname -> time added (uint64 seconds since epoch) 5152 * ... 5153 * } 5154 */ 5155 /* ARGSUSED */ 5156 static int 5157 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl) 5158 { 5159 return (dsl_dataset_get_holds(snapname, outnvl)); 5160 } 5161 5162 /* 5163 * innvl: { 5164 * snapname -> { holdname, ... } 5165 * ... 5166 * } 5167 * 5168 * outnvl: { 5169 * snapname -> error value (int32) 5170 * ... 5171 * } 5172 */ 5173 /* ARGSUSED */ 5174 static int 5175 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist) 5176 { 5177 return (dsl_dataset_user_release(holds, errlist)); 5178 } 5179 5180 /* 5181 * inputs: 5182 * zc_name name of new filesystem or snapshot 5183 * zc_value full name of old snapshot 5184 * 5185 * outputs: 5186 * zc_cookie space in bytes 5187 * zc_objset_type compressed space in bytes 5188 * zc_perm_action uncompressed space in bytes 5189 */ 5190 static int 5191 zfs_ioc_space_written(zfs_cmd_t *zc) 5192 { 5193 int error; 5194 dsl_pool_t *dp; 5195 dsl_dataset_t *new, *old; 5196 5197 error = dsl_pool_hold(zc->zc_name, FTAG, &dp); 5198 if (error != 0) 5199 return (error); 5200 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new); 5201 if (error != 0) { 5202 dsl_pool_rele(dp, FTAG); 5203 return (error); 5204 } 5205 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old); 5206 if (error != 0) { 5207 dsl_dataset_rele(new, FTAG); 5208 dsl_pool_rele(dp, FTAG); 5209 return (error); 5210 } 5211 5212 error = dsl_dataset_space_written(old, new, &zc->zc_cookie, 5213 &zc->zc_objset_type, &zc->zc_perm_action); 5214 dsl_dataset_rele(old, FTAG); 5215 dsl_dataset_rele(new, FTAG); 5216 dsl_pool_rele(dp, FTAG); 5217 return (error); 5218 } 5219 5220 /* 5221 * innvl: { 5222 * "firstsnap" -> snapshot name 5223 * } 5224 * 5225 * outnvl: { 5226 * "used" -> space in bytes 5227 * "compressed" -> compressed space in bytes 5228 * "uncompressed" -> uncompressed space in bytes 5229 * } 5230 */ 5231 static int 5232 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl) 5233 { 5234 int error; 5235 dsl_pool_t *dp; 5236 dsl_dataset_t *new, *old; 5237 char *firstsnap; 5238 uint64_t used, comp, uncomp; 5239 5240 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0) 5241 return (SET_ERROR(EINVAL)); 5242 5243 error = dsl_pool_hold(lastsnap, FTAG, &dp); 5244 if (error != 0) 5245 return (error); 5246 5247 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new); 5248 if (error != 0) { 5249 dsl_pool_rele(dp, FTAG); 5250 return (error); 5251 } 5252 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old); 5253 if (error != 0) { 5254 dsl_dataset_rele(new, FTAG); 5255 dsl_pool_rele(dp, FTAG); 5256 return (error); 5257 } 5258 5259 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp); 5260 dsl_dataset_rele(old, FTAG); 5261 dsl_dataset_rele(new, FTAG); 5262 dsl_pool_rele(dp, FTAG); 5263 fnvlist_add_uint64(outnvl, "used", used); 5264 fnvlist_add_uint64(outnvl, "compressed", comp); 5265 fnvlist_add_uint64(outnvl, "uncompressed", uncomp); 5266 return (error); 5267 } 5268 5269 /* 5270 * innvl: { 5271 * "fd" -> file descriptor to write stream to (int32) 5272 * (optional) "fromsnap" -> full snap name to send an incremental from 5273 * (optional) "largeblockok" -> (value ignored) 5274 * indicates that blocks > 128KB are permitted 5275 * (optional) "embedok" -> (value ignored) 5276 * presence indicates DRR_WRITE_EMBEDDED records are permitted 5277 * } 5278 * 5279 * outnvl is unused 5280 */ 5281 /* ARGSUSED */ 5282 static int 5283 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5284 { 5285 int error; 5286 offset_t off; 5287 char *fromname = NULL; 5288 int fd; 5289 boolean_t largeblockok; 5290 boolean_t embedok; 5291 5292 error = nvlist_lookup_int32(innvl, "fd", &fd); 5293 if (error != 0) 5294 return (SET_ERROR(EINVAL)); 5295 5296 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname); 5297 5298 largeblockok = nvlist_exists(innvl, "largeblockok"); 5299 embedok = nvlist_exists(innvl, "embedok"); 5300 5301 file_t *fp = getf(fd); 5302 if (fp == NULL) 5303 return (SET_ERROR(EBADF)); 5304 5305 off = fp->f_offset; 5306 error = dmu_send(snapname, fromname, embedok, largeblockok, 5307 fd, fp->f_vnode, &off); 5308 5309 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0) 5310 fp->f_offset = off; 5311 releasef(fd); 5312 return (error); 5313 } 5314 5315 /* 5316 * Determine approximately how large a zfs send stream will be -- the number 5317 * of bytes that will be written to the fd supplied to zfs_ioc_send_new(). 5318 * 5319 * innvl: { 5320 * (optional) "from" -> full snap or bookmark name to send an incremental 5321 * from 5322 * } 5323 * 5324 * outnvl: { 5325 * "space" -> bytes of space (uint64) 5326 * } 5327 */ 5328 static int 5329 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl) 5330 { 5331 dsl_pool_t *dp; 5332 dsl_dataset_t *tosnap; 5333 int error; 5334 char *fromname; 5335 uint64_t space; 5336 5337 error = dsl_pool_hold(snapname, FTAG, &dp); 5338 if (error != 0) 5339 return (error); 5340 5341 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap); 5342 if (error != 0) { 5343 dsl_pool_rele(dp, FTAG); 5344 return (error); 5345 } 5346 5347 error = nvlist_lookup_string(innvl, "from", &fromname); 5348 if (error == 0) { 5349 if (strchr(fromname, '@') != NULL) { 5350 /* 5351 * If from is a snapshot, hold it and use the more 5352 * efficient dmu_send_estimate to estimate send space 5353 * size using deadlists. 5354 */ 5355 dsl_dataset_t *fromsnap; 5356 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap); 5357 if (error != 0) 5358 goto out; 5359 error = dmu_send_estimate(tosnap, fromsnap, &space); 5360 dsl_dataset_rele(fromsnap, FTAG); 5361 } else if (strchr(fromname, '#') != NULL) { 5362 /* 5363 * If from is a bookmark, fetch the creation TXG of the 5364 * snapshot it was created from and use that to find 5365 * blocks that were born after it. 5366 */ 5367 zfs_bookmark_phys_t frombm; 5368 5369 error = dsl_bookmark_lookup(dp, fromname, tosnap, 5370 &frombm); 5371 if (error != 0) 5372 goto out; 5373 error = dmu_send_estimate_from_txg(tosnap, 5374 frombm.zbm_creation_txg, &space); 5375 } else { 5376 /* 5377 * from is not properly formatted as a snapshot or 5378 * bookmark 5379 */ 5380 error = SET_ERROR(EINVAL); 5381 goto out; 5382 } 5383 } else { 5384 // If estimating the size of a full send, use dmu_send_estimate 5385 error = dmu_send_estimate(tosnap, NULL, &space); 5386 } 5387 5388 fnvlist_add_uint64(outnvl, "space", space); 5389 5390 out: 5391 dsl_dataset_rele(tosnap, FTAG); 5392 dsl_pool_rele(dp, FTAG); 5393 return (error); 5394 } 5395 5396 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST]; 5397 5398 static void 5399 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5400 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5401 boolean_t log_history, zfs_ioc_poolcheck_t pool_check) 5402 { 5403 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5404 5405 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5406 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5407 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5408 ASSERT3P(vec->zvec_func, ==, NULL); 5409 5410 vec->zvec_legacy_func = func; 5411 vec->zvec_secpolicy = secpolicy; 5412 vec->zvec_namecheck = namecheck; 5413 vec->zvec_allow_log = log_history; 5414 vec->zvec_pool_check = pool_check; 5415 } 5416 5417 /* 5418 * See the block comment at the beginning of this file for details on 5419 * each argument to this function. 5420 */ 5421 static void 5422 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, 5423 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck, 5424 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist, 5425 boolean_t allow_log) 5426 { 5427 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST]; 5428 5429 ASSERT3U(ioc, >=, ZFS_IOC_FIRST); 5430 ASSERT3U(ioc, <, ZFS_IOC_LAST); 5431 ASSERT3P(vec->zvec_legacy_func, ==, NULL); 5432 ASSERT3P(vec->zvec_func, ==, NULL); 5433 5434 /* if we are logging, the name must be valid */ 5435 ASSERT(!allow_log || namecheck != NO_NAME); 5436 5437 vec->zvec_name = name; 5438 vec->zvec_func = func; 5439 vec->zvec_secpolicy = secpolicy; 5440 vec->zvec_namecheck = namecheck; 5441 vec->zvec_pool_check = pool_check; 5442 vec->zvec_smush_outnvlist = smush_outnvlist; 5443 vec->zvec_allow_log = allow_log; 5444 } 5445 5446 static void 5447 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5448 zfs_secpolicy_func_t *secpolicy, boolean_t log_history, 5449 zfs_ioc_poolcheck_t pool_check) 5450 { 5451 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5452 POOL_NAME, log_history, pool_check); 5453 } 5454 5455 static void 5456 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5457 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check) 5458 { 5459 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5460 DATASET_NAME, B_FALSE, pool_check); 5461 } 5462 5463 static void 5464 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5465 { 5466 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config, 5467 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5468 } 5469 5470 static void 5471 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5472 zfs_secpolicy_func_t *secpolicy) 5473 { 5474 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5475 NO_NAME, B_FALSE, POOL_CHECK_NONE); 5476 } 5477 5478 static void 5479 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc, 5480 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy) 5481 { 5482 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5483 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED); 5484 } 5485 5486 static void 5487 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func) 5488 { 5489 zfs_ioctl_register_dataset_read_secpolicy(ioc, func, 5490 zfs_secpolicy_read); 5491 } 5492 5493 static void 5494 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, 5495 zfs_secpolicy_func_t *secpolicy) 5496 { 5497 zfs_ioctl_register_legacy(ioc, func, secpolicy, 5498 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5499 } 5500 5501 static void 5502 zfs_ioctl_init(void) 5503 { 5504 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT, 5505 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME, 5506 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5507 5508 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY, 5509 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME, 5510 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE); 5511 5512 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS, 5513 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, 5514 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5515 5516 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW, 5517 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME, 5518 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5519 5520 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE, 5521 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME, 5522 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5523 5524 zfs_ioctl_register("create", ZFS_IOC_CREATE, 5525 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME, 5526 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5527 5528 zfs_ioctl_register("clone", ZFS_IOC_CLONE, 5529 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME, 5530 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5531 5532 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS, 5533 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME, 5534 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5535 5536 zfs_ioctl_register("hold", ZFS_IOC_HOLD, 5537 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME, 5538 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5539 zfs_ioctl_register("release", ZFS_IOC_RELEASE, 5540 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME, 5541 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5542 5543 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS, 5544 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, 5545 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5546 5547 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK, 5548 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, 5549 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE); 5550 5551 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK, 5552 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME, 5553 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5554 5555 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS, 5556 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME, 5557 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE); 5558 5559 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS, 5560 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks, 5561 POOL_NAME, 5562 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE); 5563 5564 /* IOCTLS that use the legacy function signature */ 5565 5566 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze, 5567 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY); 5568 5569 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create, 5570 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5571 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN, 5572 zfs_ioc_pool_scan); 5573 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE, 5574 zfs_ioc_pool_upgrade); 5575 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD, 5576 zfs_ioc_vdev_add); 5577 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE, 5578 zfs_ioc_vdev_remove); 5579 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE, 5580 zfs_ioc_vdev_set_state); 5581 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH, 5582 zfs_ioc_vdev_attach); 5583 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH, 5584 zfs_ioc_vdev_detach); 5585 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH, 5586 zfs_ioc_vdev_setpath); 5587 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU, 5588 zfs_ioc_vdev_setfru); 5589 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS, 5590 zfs_ioc_pool_set_props); 5591 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT, 5592 zfs_ioc_vdev_split); 5593 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID, 5594 zfs_ioc_pool_reguid); 5595 5596 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS, 5597 zfs_ioc_pool_configs, zfs_secpolicy_none); 5598 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT, 5599 zfs_ioc_pool_tryimport, zfs_secpolicy_config); 5600 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT, 5601 zfs_ioc_inject_fault, zfs_secpolicy_inject); 5602 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT, 5603 zfs_ioc_clear_fault, zfs_secpolicy_inject); 5604 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT, 5605 zfs_ioc_inject_list_next, zfs_secpolicy_inject); 5606 5607 /* 5608 * pool destroy, and export don't log the history as part of 5609 * zfsdev_ioctl, but rather zfs_ioc_pool_export 5610 * does the logging of those commands. 5611 */ 5612 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy, 5613 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5614 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export, 5615 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE); 5616 5617 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats, 5618 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5619 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props, 5620 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE); 5621 5622 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log, 5623 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED); 5624 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME, 5625 zfs_ioc_dsobj_to_dsname, 5626 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED); 5627 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY, 5628 zfs_ioc_pool_get_history, 5629 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED); 5630 5631 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import, 5632 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5633 5634 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear, 5635 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE); 5636 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen, 5637 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED); 5638 5639 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN, 5640 zfs_ioc_space_written); 5641 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS, 5642 zfs_ioc_objset_recvd_props); 5643 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ, 5644 zfs_ioc_next_obj); 5645 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL, 5646 zfs_ioc_get_fsacl); 5647 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS, 5648 zfs_ioc_objset_stats); 5649 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS, 5650 zfs_ioc_objset_zplprops); 5651 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT, 5652 zfs_ioc_dataset_list_next); 5653 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT, 5654 zfs_ioc_snapshot_list_next); 5655 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS, 5656 zfs_ioc_send_progress); 5657 5658 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF, 5659 zfs_ioc_diff, zfs_secpolicy_diff); 5660 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS, 5661 zfs_ioc_obj_to_stats, zfs_secpolicy_diff); 5662 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH, 5663 zfs_ioc_obj_to_path, zfs_secpolicy_diff); 5664 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE, 5665 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one); 5666 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY, 5667 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many); 5668 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND, 5669 zfs_ioc_send, zfs_secpolicy_send); 5670 5671 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop, 5672 zfs_secpolicy_none); 5673 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy, 5674 zfs_secpolicy_destroy); 5675 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename, 5676 zfs_secpolicy_rename); 5677 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv, 5678 zfs_secpolicy_recv); 5679 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote, 5680 zfs_secpolicy_promote); 5681 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP, 5682 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop); 5683 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl, 5684 zfs_secpolicy_set_fsacl); 5685 5686 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share, 5687 zfs_secpolicy_share, POOL_CHECK_NONE); 5688 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl, 5689 zfs_secpolicy_smb_acl, POOL_CHECK_NONE); 5690 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE, 5691 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade, 5692 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5693 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT, 5694 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, 5695 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY); 5696 } 5697 5698 int 5699 pool_status_check(const char *name, zfs_ioc_namecheck_t type, 5700 zfs_ioc_poolcheck_t check) 5701 { 5702 spa_t *spa; 5703 int error; 5704 5705 ASSERT(type == POOL_NAME || type == DATASET_NAME); 5706 5707 if (check & POOL_CHECK_NONE) 5708 return (0); 5709 5710 error = spa_open(name, &spa, FTAG); 5711 if (error == 0) { 5712 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa)) 5713 error = SET_ERROR(EAGAIN); 5714 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa)) 5715 error = SET_ERROR(EROFS); 5716 spa_close(spa, FTAG); 5717 } 5718 return (error); 5719 } 5720 5721 /* 5722 * Find a free minor number. 5723 */ 5724 minor_t 5725 zfsdev_minor_alloc(void) 5726 { 5727 static minor_t last_minor; 5728 minor_t m; 5729 5730 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 5731 5732 for (m = last_minor + 1; m != last_minor; m++) { 5733 if (m > ZFSDEV_MAX_MINOR) 5734 m = 1; 5735 if (ddi_get_soft_state(zfsdev_state, m) == NULL) { 5736 last_minor = m; 5737 return (m); 5738 } 5739 } 5740 5741 return (0); 5742 } 5743 5744 static int 5745 zfs_ctldev_init(dev_t *devp) 5746 { 5747 minor_t minor; 5748 zfs_soft_state_t *zs; 5749 5750 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 5751 ASSERT(getminor(*devp) == 0); 5752 5753 minor = zfsdev_minor_alloc(); 5754 if (minor == 0) 5755 return (SET_ERROR(ENXIO)); 5756 5757 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) 5758 return (SET_ERROR(EAGAIN)); 5759 5760 *devp = makedevice(getemajor(*devp), minor); 5761 5762 zs = ddi_get_soft_state(zfsdev_state, minor); 5763 zs->zss_type = ZSST_CTLDEV; 5764 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data); 5765 5766 return (0); 5767 } 5768 5769 static void 5770 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor) 5771 { 5772 ASSERT(MUTEX_HELD(&zfsdev_state_lock)); 5773 5774 zfs_onexit_destroy(zo); 5775 ddi_soft_state_free(zfsdev_state, minor); 5776 } 5777 5778 void * 5779 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which) 5780 { 5781 zfs_soft_state_t *zp; 5782 5783 zp = ddi_get_soft_state(zfsdev_state, minor); 5784 if (zp == NULL || zp->zss_type != which) 5785 return (NULL); 5786 5787 return (zp->zss_data); 5788 } 5789 5790 static int 5791 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr) 5792 { 5793 int error = 0; 5794 5795 if (getminor(*devp) != 0) 5796 return (zvol_open(devp, flag, otyp, cr)); 5797 5798 /* This is the control device. Allocate a new minor if requested. */ 5799 if (flag & FEXCL) { 5800 mutex_enter(&zfsdev_state_lock); 5801 error = zfs_ctldev_init(devp); 5802 mutex_exit(&zfsdev_state_lock); 5803 } 5804 5805 return (error); 5806 } 5807 5808 static int 5809 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr) 5810 { 5811 zfs_onexit_t *zo; 5812 minor_t minor = getminor(dev); 5813 5814 if (minor == 0) 5815 return (0); 5816 5817 mutex_enter(&zfsdev_state_lock); 5818 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV); 5819 if (zo == NULL) { 5820 mutex_exit(&zfsdev_state_lock); 5821 return (zvol_close(dev, flag, otyp, cr)); 5822 } 5823 zfs_ctldev_destroy(zo, minor); 5824 mutex_exit(&zfsdev_state_lock); 5825 5826 return (0); 5827 } 5828 5829 static int 5830 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 5831 { 5832 zfs_cmd_t *zc; 5833 uint_t vecnum; 5834 int error, rc, len; 5835 minor_t minor = getminor(dev); 5836 const zfs_ioc_vec_t *vec; 5837 char *saved_poolname = NULL; 5838 nvlist_t *innvl = NULL; 5839 5840 if (minor != 0 && 5841 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL) 5842 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp)); 5843 5844 vecnum = cmd - ZFS_IOC_FIRST; 5845 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip)); 5846 5847 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0])) 5848 return (SET_ERROR(EINVAL)); 5849 vec = &zfs_ioc_vec[vecnum]; 5850 5851 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP); 5852 5853 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag); 5854 if (error != 0) { 5855 error = SET_ERROR(EFAULT); 5856 goto out; 5857 } 5858 5859 zc->zc_iflags = flag & FKIOCTL; 5860 if (zc->zc_nvlist_src_size != 0) { 5861 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, 5862 zc->zc_iflags, &innvl); 5863 if (error != 0) 5864 goto out; 5865 } 5866 5867 /* 5868 * Ensure that all pool/dataset names are valid before we pass down to 5869 * the lower layers. 5870 */ 5871 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0'; 5872 switch (vec->zvec_namecheck) { 5873 case POOL_NAME: 5874 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0) 5875 error = SET_ERROR(EINVAL); 5876 else 5877 error = pool_status_check(zc->zc_name, 5878 vec->zvec_namecheck, vec->zvec_pool_check); 5879 break; 5880 5881 case DATASET_NAME: 5882 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0) 5883 error = SET_ERROR(EINVAL); 5884 else 5885 error = pool_status_check(zc->zc_name, 5886 vec->zvec_namecheck, vec->zvec_pool_check); 5887 break; 5888 5889 case NO_NAME: 5890 break; 5891 } 5892 5893 5894 if (error == 0 && !(flag & FKIOCTL)) 5895 error = vec->zvec_secpolicy(zc, innvl, cr); 5896 5897 if (error != 0) 5898 goto out; 5899 5900 /* legacy ioctls can modify zc_name */ 5901 len = strcspn(zc->zc_name, "/@#") + 1; 5902 saved_poolname = kmem_alloc(len, KM_SLEEP); 5903 (void) strlcpy(saved_poolname, zc->zc_name, len); 5904 5905 if (vec->zvec_func != NULL) { 5906 nvlist_t *outnvl; 5907 int puterror = 0; 5908 spa_t *spa; 5909 nvlist_t *lognv = NULL; 5910 5911 ASSERT(vec->zvec_legacy_func == NULL); 5912 5913 /* 5914 * Add the innvl to the lognv before calling the func, 5915 * in case the func changes the innvl. 5916 */ 5917 if (vec->zvec_allow_log) { 5918 lognv = fnvlist_alloc(); 5919 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL, 5920 vec->zvec_name); 5921 if (!nvlist_empty(innvl)) { 5922 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL, 5923 innvl); 5924 } 5925 } 5926 5927 outnvl = fnvlist_alloc(); 5928 error = vec->zvec_func(zc->zc_name, innvl, outnvl); 5929 5930 if (error == 0 && vec->zvec_allow_log && 5931 spa_open(zc->zc_name, &spa, FTAG) == 0) { 5932 if (!nvlist_empty(outnvl)) { 5933 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL, 5934 outnvl); 5935 } 5936 (void) spa_history_log_nvl(spa, lognv); 5937 spa_close(spa, FTAG); 5938 } 5939 fnvlist_free(lognv); 5940 5941 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) { 5942 int smusherror = 0; 5943 if (vec->zvec_smush_outnvlist) { 5944 smusherror = nvlist_smush(outnvl, 5945 zc->zc_nvlist_dst_size); 5946 } 5947 if (smusherror == 0) 5948 puterror = put_nvlist(zc, outnvl); 5949 } 5950 5951 if (puterror != 0) 5952 error = puterror; 5953 5954 nvlist_free(outnvl); 5955 } else { 5956 error = vec->zvec_legacy_func(zc); 5957 } 5958 5959 out: 5960 nvlist_free(innvl); 5961 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag); 5962 if (error == 0 && rc != 0) 5963 error = SET_ERROR(EFAULT); 5964 if (error == 0 && vec->zvec_allow_log) { 5965 char *s = tsd_get(zfs_allow_log_key); 5966 if (s != NULL) 5967 strfree(s); 5968 (void) tsd_set(zfs_allow_log_key, saved_poolname); 5969 } else { 5970 if (saved_poolname != NULL) 5971 strfree(saved_poolname); 5972 } 5973 5974 kmem_free(zc, sizeof (zfs_cmd_t)); 5975 return (error); 5976 } 5977 5978 static int 5979 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd) 5980 { 5981 if (cmd != DDI_ATTACH) 5982 return (DDI_FAILURE); 5983 5984 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0, 5985 DDI_PSEUDO, 0) == DDI_FAILURE) 5986 return (DDI_FAILURE); 5987 5988 zfs_dip = dip; 5989 5990 ddi_report_dev(dip); 5991 5992 return (DDI_SUCCESS); 5993 } 5994 5995 static int 5996 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd) 5997 { 5998 if (spa_busy() || zfs_busy() || zvol_busy()) 5999 return (DDI_FAILURE); 6000 6001 if (cmd != DDI_DETACH) 6002 return (DDI_FAILURE); 6003 6004 zfs_dip = NULL; 6005 6006 ddi_prop_remove_all(dip); 6007 ddi_remove_minor_node(dip, NULL); 6008 6009 return (DDI_SUCCESS); 6010 } 6011 6012 /*ARGSUSED*/ 6013 static int 6014 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result) 6015 { 6016 switch (infocmd) { 6017 case DDI_INFO_DEVT2DEVINFO: 6018 *result = zfs_dip; 6019 return (DDI_SUCCESS); 6020 6021 case DDI_INFO_DEVT2INSTANCE: 6022 *result = (void *)0; 6023 return (DDI_SUCCESS); 6024 } 6025 6026 return (DDI_FAILURE); 6027 } 6028 6029 /* 6030 * OK, so this is a little weird. 6031 * 6032 * /dev/zfs is the control node, i.e. minor 0. 6033 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0. 6034 * 6035 * /dev/zfs has basically nothing to do except serve up ioctls, 6036 * so most of the standard driver entry points are in zvol.c. 6037 */ 6038 static struct cb_ops zfs_cb_ops = { 6039 zfsdev_open, /* open */ 6040 zfsdev_close, /* close */ 6041 zvol_strategy, /* strategy */ 6042 nodev, /* print */ 6043 zvol_dump, /* dump */ 6044 zvol_read, /* read */ 6045 zvol_write, /* write */ 6046 zfsdev_ioctl, /* ioctl */ 6047 nodev, /* devmap */ 6048 nodev, /* mmap */ 6049 nodev, /* segmap */ 6050 nochpoll, /* poll */ 6051 ddi_prop_op, /* prop_op */ 6052 NULL, /* streamtab */ 6053 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */ 6054 CB_REV, /* version */ 6055 nodev, /* async read */ 6056 nodev, /* async write */ 6057 }; 6058 6059 static struct dev_ops zfs_dev_ops = { 6060 DEVO_REV, /* version */ 6061 0, /* refcnt */ 6062 zfs_info, /* info */ 6063 nulldev, /* identify */ 6064 nulldev, /* probe */ 6065 zfs_attach, /* attach */ 6066 zfs_detach, /* detach */ 6067 nodev, /* reset */ 6068 &zfs_cb_ops, /* driver operations */ 6069 NULL, /* no bus operations */ 6070 NULL, /* power */ 6071 ddi_quiesce_not_needed, /* quiesce */ 6072 }; 6073 6074 static struct modldrv zfs_modldrv = { 6075 &mod_driverops, 6076 "ZFS storage pool", 6077 &zfs_dev_ops 6078 }; 6079 6080 static struct modlinkage modlinkage = { 6081 MODREV_1, 6082 (void *)&zfs_modlfs, 6083 (void *)&zfs_modldrv, 6084 NULL 6085 }; 6086 6087 static void 6088 zfs_allow_log_destroy(void *arg) 6089 { 6090 char *poolname = arg; 6091 strfree(poolname); 6092 } 6093 6094 int 6095 _init(void) 6096 { 6097 int error; 6098 6099 spa_init(FREAD | FWRITE); 6100 zfs_init(); 6101 zvol_init(); 6102 zfs_ioctl_init(); 6103 6104 if ((error = mod_install(&modlinkage)) != 0) { 6105 zvol_fini(); 6106 zfs_fini(); 6107 spa_fini(); 6108 return (error); 6109 } 6110 6111 tsd_create(&zfs_fsyncer_key, NULL); 6112 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 6113 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy); 6114 6115 error = ldi_ident_from_mod(&modlinkage, &zfs_li); 6116 ASSERT(error == 0); 6117 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL); 6118 6119 return (0); 6120 } 6121 6122 int 6123 _fini(void) 6124 { 6125 int error; 6126 6127 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled) 6128 return (SET_ERROR(EBUSY)); 6129 6130 if ((error = mod_remove(&modlinkage)) != 0) 6131 return (error); 6132 6133 zvol_fini(); 6134 zfs_fini(); 6135 spa_fini(); 6136 if (zfs_nfsshare_inited) 6137 (void) ddi_modclose(nfs_mod); 6138 if (zfs_smbshare_inited) 6139 (void) ddi_modclose(smbsrv_mod); 6140 if (zfs_nfsshare_inited || zfs_smbshare_inited) 6141 (void) ddi_modclose(sharefs_mod); 6142 6143 tsd_destroy(&zfs_fsyncer_key); 6144 ldi_ident_release(zfs_li); 6145 zfs_li = NULL; 6146 mutex_destroy(&zfs_share_lock); 6147 6148 return (error); 6149 } 6150 6151 int 6152 _info(struct modinfo *modinfop) 6153 { 6154 return (mod_info(&modlinkage, modinfop)); 6155 } 6156